Realclearpolitics is interesting.

[Cali Coug]

I like it because it assembles so many polls in one easy to find place.

I am a bit confused, though, as to why the RCP average (which is just an average of certain polls- I don't even know how they pick the polls to average) is taken so seriously.

There is no statistical reason why the average of all of the polls should be more accurate than any individual poll. So what gives?

[YOhio]

Why would you be so cruel in bursting myboynoah's balloon? You know how much he loves RCP average. What's next? Making fun of InTrade?

[Cali Coug]

Originally Posted by YOhio

Why would you be so cruel in bursting myboynoah's balloon? You know how much he loves RCP average. What's next? Making fun of InTrade?

I already did that too...

[jay santos]

Originally Posted by Cali Coug

I like it because it assembles so many polls in one easy to find place.

I am a bit confused, though, as to why the RCP average (which is just an average of certain polls- I don't even know how they pick the polls to average) is taken so seriously.

There is no statistical reason why the average of all of the polls should be more accurate than any individual poll. So what gives?

Sure there's a statistical reason an average of polls is more accurate than an individual poll.

Same concept of a mutual fund reduces risk compared to buying one stock.

[myboynoah]

Originally Posted by YOhio

Why would you be so cruel in bursting myboynoah's balloon? You know how much he loves RCP average. What's next? Making fun of InTrade?

Oh, so what is this, an intervention? I'm not going anywhere. I don't have a problem. It is you who has the problem.

The only importance I've placed on the RCP average is that for some inexplicable reason some news services place importance upon it. I agree with Cali, very odd. Nonetheless, RealClearPolitics is a very intriguing website.

Had you gotten behind Romney early on Intrade, you'd be a rich man now. Sadly, that ship has sailed.

[All-American]

Originally Posted by myboynoah

Oh, so what is this, an intervention?

Sounds more like an imposition.

[Cali Coug]

Originally Posted by jay santos

Sure there's a statistical reason an average of polls is more accurate than an individual poll.

Same concept of a mutual fund reduces risk compared to buying one stock.

Come on, Jay. You know better than that. What is the margin of error of the polls being combined? Is it the same in each? If not, how do you weight one over the other? Can you be certain each used the same methodology? Do you have any idea if the polls are at all comparable?

In a mutual fund, you are hedging against the chance that one stock fails by diversifying the portfolio. In other words, you are going to get some crap stocks and some good stocks (you hope, though that isn't guaranteed either). Why would you expect a combination of polls to be more "accurate?" It may hedge against bad polls, and it may include some good polls. Or they could all be crap, or they could all be good, and averaging them could dilute the influence of the "right" poll.

[Jeff Lebowski]

Originally Posted by Cali Coug

Come on, Jay. You know better than that. What is the margin of error of the polls being combined? Is it the same in each? If not, how do you weight one over the other? Can you be certain each used the same methodology? Do you have any idea if the polls are at all comparable?

In a mutual fund, you are hedging against the chance that one stock fails by diversifying the portfolio. In other words, you are going to get some crap stocks and some good stocks (you hope, though that isn't guaranteed either). Why would you expect a combination of polls to be more "accurate?" It may hedge against bad polls, and it may include some good polls. Or they could all be crap, or they could all be good, and averaging them could dilute the influence of the "right" poll.

You are correct that a more accurate approach would involve some sort of weighting factors when the polls are combined. But one of the factors leading to a margin of error is the sample size. By pooling the polls the same size increases. Common sense would argue that the combined polls are likely more accurate than the individual polls.

And I have seen research indicating that the average of independent predictions are more accurate than individual predictions. These were predictions made by computer models, not polls.

And scientific arguments aside, I think there is value in seeing a summary of all of the polls in one spot.

[jay santos]

Originally Posted by Cali Coug

Come on, Jay. You know better than that. What is the margin of error of the polls being combined? Is it the same in each? If not, how do you weight one over the other? Can you be certain each used the same methodology? Do you have any idea if the polls are at all comparable?

In a mutual fund, you are hedging against the chance that one stock fails by diversifying the portfolio. In other words, you are going to get some crap stocks and some good stocks (you hope, though that isn't guaranteed either). Why would you expect a combination of polls to be more "accurate?" It may hedge against bad polls, and it may include some good polls. Or they could all be crap, or they could all be good, and averaging them could dilute the influence of the "right" poll.

If you knew which one was right, you wouldn't have to have an average, but you don't. Just like if you knew which stock was "right", you wouldn't need to dilute it by averaging it with other stocks in a mutual fund.

If the polls have reasonably similar methodology and all have reasonably credibility when it comes to accuracy (I don't know if they do but assume they do), then averaging them gives a better chance for accuracy than any one poll by itself.

[Cali Coug]

Originally Posted by jay santos

If you knew which one was right, you wouldn't have to have an average, but you don't. Just like if you knew which stock was "right", you wouldn't need to dilute it by averaging it with other stocks in a mutual fund.

If the polls have reasonably similar methodology and all have reasonably credibility when it comes to accuracy (I don't know if they do but assume they do), then averaging them gives a better chance for accuracy than any one poll by itself.

They don't all have comparable credibility, and they don't all use the same methodology. And there are surveys done to identify which pollsters tend to get it right more than the others (Rasmussen is nearly always hovering around the top, for example).