What should be done right now regarding polls added to the database ?

[Tender Branson]

Inks and I have discussed recently how to enter polls to the database in the next few months.

I favor this:

General election polls should be entered to the database using the leader of the most recent GOP primary poll in each state. For example: Santorum leads in OH's primary, he should be entered as Obama's opponent. If Romney wins OH on Super Tuesday, we pick Romney as Obama's opponent.

Inks favors this:

We should establish a current "frontrunner", say Romney, and use him for every poll until we know the eventual nominee.

...

What do you say ?

[Inks.LWC Supports Chuck Hagel]

The problem with using who leads in the polls state-to-state is that it's not representational as to who will actually face Obama.  Gingrich has a minute chance of winning the nomination, but under the Branson Option, we'd be using Gingrich in match-ups against Obama in Georgia.

And I need to correct something in the Inks option.  We establish a frontrunner, but if the frontrunner changes, we change who we use in each poll.  So, if in January, we said it was Romney, we'd all use Romney for all polls.  Now, if we decide it's Santorum, we use Santorum in all the polls.  But if Romney has a great showing on Super Tuesday and becomes the frontrunner, we'd go back to him.

That method at least keeps national consistency, and seems to be the closest to the method Dave has had in mind before.

[Inks.LWC Supports Chuck Hagel]

Another problem with doing it state-by-state is what do we do when states have conflicting recent polls or no recent polls at all?

[Tender Branson]

So, if we use the Inks option in the next months, who do we designate as the "front runner" ?

Santorum because he leads the national polls, or Romney because he leads among delegates ?

[Χahar]

I'm inclined to agree with Inks here; putting in different candidates based on the state will exaggerate Republican support.

So, if we use the Inks option in the next months, who do we designate as the "front runner" ?

Santorum because he leads the national polls, or Romney because he leads among delegates ?

National polls are meaningless; you would have to go by delegates.

[Gustaf]

Tender's method makes no sense, since it doesn't display anything. The race won't be Obama versus whoever Republican won what primary. It'll be Obama against the Republican nominee. Thus, we should make our best guess as to who that will be and go by that.

Right now I think the best guess is Romney so we should stick to that for the time being.

[Californian Tony]

I have to agree with Inks. Unless the GOP decides to pull a Whig 1836, there will be the same candidate in all States.

[© Tweed the Younger]

as much as I hate to choose the Inks option, I choose the Inks option.

[Nym90]

I have to agree with Inks. Unless the GOP decides to pull a Whig 1836, there will be the same candidate in all States.

I'm kind of surprised no one has mentioned that as one possible solution if the GOP ends up with a brokered convention. They could put Romney, Santorum or Gingrich each on the ballot for the general election in the states that they won in the primaries, in the hopes that collectively they could achieve a majority of the electoral votes, throwing the election to what will probably still be a GOP majority in the House (or the GOP electors themselves could all agree to support one candidate when the Electoral College votes).

The risk would be that the public would frown on this strategy and punish the GOP for pursuing it, but there's no requirement whatsoever that a party have the same nominee in every state.

The Dems could have tried the same thing in 2008 with Clinton and Obama.

[Beet]

Another option is that we enter the average of all four Republicans, or however many Republicans are polled, versus the Obama average. For example, if a poll shows

Obama 48 Romney 48
Obama 49 Santorum 46
Obama 47 Paul 45
Obama 51 Gingrich 41
----------------------------
(48+49+47+51) / 4 = 195 / 4 = 48.75 = 49 Obama
(48+46+45+41) / 4 = 180 / 4 = 45 Republican

It requires a bit more calculation, but it's more accurate to get a time series comparing across months. By the Inks method, if Romney leads in January and Santorum takes over in February, let's say Santorum polls much more weakly than Romney, it would look as if the country suddenly shifted to Obama when in fact, only the GOP frontrunner changed. Plus, what if the frontrunner changes every few days, or there is no frontrunner? Then it'd be chaos. My method is still somewhat subject to this (based on when people drop out) but it's more stable. It does require a bit more calculation, but it only takes a minute or do to do an average. Just type in the formula to Google if you don't have a calculator.

[Inks.LWC Supports Chuck Hagel]

The problem with that is that not all posters include all 4 candidates, and we're trying to be as consistent as possible.

Also, Gingrich typically does worse than the 2 who have a chance of actually being the nominee.  Including him artificially lower's the GOP's actual chances in that state.

[Beet]

The problem with that is that not all posters include all 4 candidates, and we're trying to be as consistent as possible.

Also, Gingrich typically does worse than the 2 who have a chance of actually being the nominee.  Including him artificially lower's the GOP's actual chances in that state.

True, but your system isn't consistent either when the leader is frequently changing. And consistency across pollsters is kind of a silly goal, since all pollsters have different methodologies. One other advantage of my system is that candidates (like Gingrich) who are less likely to win the nomination are less likely to be polled, so they'll end up being weighted less in the overall scheme of things. Arguably, the decision of who to poll reflects the pollster's judgment about who is worthy of consideration, just like the phrasing of the question and other aspects of methodology reflect the pollster's judgment about how to tease out the most accurate result.

[Inks.LWC Supports Chuck Hagel]

The problem with that is that not all posters include all 4 candidates, and we're trying to be as consistent as possible.

Also, Gingrich typically does worse than the 2 who have a chance of actually being the nominee.  Including him artificially lower's the GOP's actual chances in that state.

True, but your system isn't consistent either when the leader is frequently changing. And consistency across pollsters is kind of a silly goal, since all pollsters have different methodologies. One other advantage of my system is that candidates (like Gingrich) who are less likely to win the nomination are less likely to be polled, so they'll end up being weighted less in the overall scheme of things. Arguably, the decision of who to poll reflects the pollster's judgment about who is worthy of consideration, just like the phrasing of the question and other aspects of methodology reflect the pollster's judgment about how to tease out the most accurate result.

Ideally, I don't think the frontrunner is changing.  I'm not basing my system off of who's leading in the latest national poll, but who is likeliest to win in the end.  Ultimately, any system is subjective, though.