Total Posts:4|Showing Posts:1-4
Jump to topic:

Reading Polls

SolonKR
Posts: 4,039
Add as Friend
Challenge to a Debate
Send a Message
11/3/2016 5:56:25 AM
Posted: 1 month ago
This isn't a political post per se; it's just a guide to help people know how exactly they should be reading the polls, as there are a lot of them, and people often misinterpret the results. It's for fun; not for advocating for a specific candidate.

It's important to distinguish between the two main types of polls: traditional polls and tracking polls. Traditional polls call up (or contact online, or otherwise) a random sample of voters, and then (each in their own way) balance that sample against the size of groups in the general electorate--the reason you see some polls polling more Democrats, for example, is that there are just more people in the country who identify as Democrat than Republican. While some of these polls are more representative than others (see FiveThirtyEight for analysis of individual pollsters and their methods), most people associate this model with polling in general.

However, tracking polls are completely different. Tracking polls (again using varied methodologies) assemble a sample group, and follow the opinions of that one group of people (they also balance that sample against the relative size of groups in the general electorate). For this reason, they are typically not as good as normal polls at representing the actual state of the race (it's not a random sample), but they're fantastic at portraying momentum, because they can look at how individual voters are changing their mind.

Now, this is where things get complicated. Since all polls weigh their responses differently, you can get really, reaaaaaaaaaaally wacky results. For example, one single 19-year old black man's support of Trump in the USC/LA Times tracking poll moves the poll results for Trump up an entire percentage point on its own, and puts his support among black voters in the poll into the double digits (lol no), even though he's one of 3000 panelists (http://www.nytimes.com...).

What does this mean? It means, quite simply, that some polls are better than others. It is a terrible, terrible idea to take the results of any one single poll and hold it up as evidence of a Clinton landslide (Clinton +6 is one result in RCP from one of the most reliable recent polls with a huuuuuuuuge sample size http://www.realclearpolitics.com...) or a Trump triumph (Trump +6 in the USC/LA Times poll http://www.realclearpolitics.com...).

Looking at RCP is one way to gauge the polls--aggregates mean that polling errors on the whole have less effect. However, if you really want to have an accurate picture of the race, I highly recommend 538 (http://projects.fivethirtyeight.com...) as vastly superior to RCP. It actually takes into account the merits and biases of each individual poll, and adjusts/weighs them accordingly. I realize that some will not like this solely on the basis that Clinton has a slightly wider margin on 538 (3.4%) than on RCP (1.9%), but at the very least, just bear in mind that a poll by itself is pretty much nothing. Look at the whole enchilada (I have a thing for mixed metaphors now #sorrynotsorry).
SO to Bailey, the love of my life <3
Vaarka
Posts: 7,533
Add as Friend
Challenge to a Debate
Send a Message
11/3/2016 12:38:50 PM
Posted: 1 month ago
At 11/3/2016 5:56:25 AM, SolonKR wrote:
This isn't a political post per se; it's just a guide to help people know how exactly they should be reading the polls, as there are a lot of them, and people often misinterpret the results. It's for fun; not for advocating for a specific candidate.

It's important to distinguish between the two main types of polls: traditional polls and tracking polls. Traditional polls call up (or contact online, or otherwise) a random sample of voters, and then (each in their own way) balance that sample against the size of groups in the general electorate--the reason you see some polls polling more Democrats, for example, is that there are just more people in the country who identify as Democrat than Republican. While some of these polls are more representative than others (see FiveThirtyEight for analysis of individual pollsters and their methods), most people associate this model with polling in general.

However, tracking polls are completely different. Tracking polls (again using varied methodologies) assemble a sample group, and follow the opinions of that one group of people (they also balance that sample against the relative size of groups in the general electorate). For this reason, they are typically not as good as normal polls at representing the actual state of the race (it's not a random sample), but they're fantastic at portraying momentum, because they can look at how individual voters are changing their mind.

Now, this is where things get complicated. Since all polls weigh their responses differently, you can get really, reaaaaaaaaaaally wacky results. For example, one single 19-year old black man's support of Trump in the USC/LA Times tracking poll moves the poll results for Trump up an entire percentage point on its own, and puts his support among black voters in the poll into the double digits (lol no), even though he's one of 3000 panelists (http://www.nytimes.com...).

What does this mean? It means, quite simply, that some polls are better than others. It is a terrible, terrible idea to take the results of any one single poll and hold it up as evidence of a Clinton landslide (Clinton +6 is one result in RCP from one of the most reliable recent polls with a huuuuuuuuge sample size http://www.realclearpolitics.com...) or a Trump triumph (Trump +6 in the USC/LA Times poll http://www.realclearpolitics.com...).

Looking at RCP is one way to gauge the polls--aggregates mean that polling errors on the whole have less effect. However, if you really want to have an accurate picture of the race, I highly recommend 538 (http://projects.fivethirtyeight.com...) as vastly superior to RCP. It actually takes into account the merits and biases of each individual poll, and adjusts/weighs them accordingly. I realize that some will not like this solely on the basis that Clinton has a slightly wider margin on 538 (3.4%) than on RCP (1.9%), but at the very least, just bear in mind that a poll by itself is pretty much nothing. Look at the whole enchilada (I have a thing for mixed metaphors now #sorrynotsorry).

So what you're saying here is that polls aren't reliable?
You're probably thinking right now "haha I'm a genius". Well you're not -Valkrin

inferno: "I don't know, are you attracted to women?"
ButterCatX: "No, Vaarka is mine!"

All hail scum Vaarka, wielder of the bastard sword, smiter of nations, destroyer of spiders -VOT

"Vaarka, I've been thinking about this for a long time now," (pulls out small box made of macaroni) "W-will you be my noodle buddy?" -Kirigaya
TBR
Posts: 9,991
Add as Friend
Challenge to a Debate
Send a Message
11/3/2016 2:01:55 PM
Posted: 1 month ago
At 11/3/2016 12:38:50 PM, Vaarka wrote:
At 11/3/2016 5:56:25 AM, SolonKR wrote:
This isn't a political post per se; it's just a guide to help people know how exactly they should be reading the polls, as there are a lot of them, and people often misinterpret the results. It's for fun; not for advocating for a specific candidate.

It's important to distinguish between the two main types of polls: traditional polls and tracking polls. Traditional polls call up (or contact online, or otherwise) a random sample of voters, and then (each in their own way) balance that sample against the size of groups in the general electorate--the reason you see some polls polling more Democrats, for example, is that there are just more people in the country who identify as Democrat than Republican. While some of these polls are more representative than others (see FiveThirtyEight for analysis of individual pollsters and their methods), most people associate this model with polling in general.

However, tracking polls are completely different. Tracking polls (again using varied methodologies) assemble a sample group, and follow the opinions of that one group of people (they also balance that sample against the relative size of groups in the general electorate). For this reason, they are typically not as good as normal polls at representing the actual state of the race (it's not a random sample), but they're fantastic at portraying momentum, because they can look at how individual voters are changing their mind.

Now, this is where things get complicated. Since all polls weigh their responses differently, you can get really, reaaaaaaaaaaally wacky results. For example, one single 19-year old black man's support of Trump in the USC/LA Times tracking poll moves the poll results for Trump up an entire percentage point on its own, and puts his support among black voters in the poll into the double digits (lol no), even though he's one of 3000 panelists (http://www.nytimes.com...).

What does this mean? It means, quite simply, that some polls are better than others. It is a terrible, terrible idea to take the results of any one single poll and hold it up as evidence of a Clinton landslide (Clinton +6 is one result in RCP from one of the most reliable recent polls with a huuuuuuuuge sample size http://www.realclearpolitics.com...) or a Trump triumph (Trump +6 in the USC/LA Times poll http://www.realclearpolitics.com...).

Looking at RCP is one way to gauge the polls--aggregates mean that polling errors on the whole have less effect. However, if you really want to have an accurate picture of the race, I highly recommend 538 (http://projects.fivethirtyeight.com...) as vastly superior to RCP. It actually takes into account the merits and biases of each individual poll, and adjusts/weighs them accordingly. I realize that some will not like this solely on the basis that Clinton has a slightly wider margin on 538 (3.4%) than on RCP (1.9%), but at the very least, just bear in mind that a poll by itself is pretty much nothing. Look at the whole enchilada (I have a thing for mixed metaphors now #sorrynotsorry).

So what you're saying here is that polls aren't reliable?

Polls are quite accurate. He is giving some good information, but is not saying that polling does not work.
Vaarka
Posts: 7,533
Add as Friend
Challenge to a Debate
Send a Message
11/3/2016 2:47:47 PM
Posted: 1 month ago
At 11/3/2016 2:01:55 PM, TBR wrote:
At 11/3/2016 12:38:50 PM, Vaarka wrote:
At 11/3/2016 5:56:25 AM, SolonKR wrote:
This isn't a political post per se; it's just a guide to help people know how exactly they should be reading the polls, as there are a lot of them, and people often misinterpret the results. It's for fun; not for advocating for a specific candidate.

It's important to distinguish between the two main types of polls: traditional polls and tracking polls. Traditional polls call up (or contact online, or otherwise) a random sample of voters, and then (each in their own way) balance that sample against the size of groups in the general electorate--the reason you see some polls polling more Democrats, for example, is that there are just more people in the country who identify as Democrat than Republican. While some of these polls are more representative than others (see FiveThirtyEight for analysis of individual pollsters and their methods), most people associate this model with polling in general.

However, tracking polls are completely different. Tracking polls (again using varied methodologies) assemble a sample group, and follow the opinions of that one group of people (they also balance that sample against the relative size of groups in the general electorate). For this reason, they are typically not as good as normal polls at representing the actual state of the race (it's not a random sample), but they're fantastic at portraying momentum, because they can look at how individual voters are changing their mind.

Now, this is where things get complicated. Since all polls weigh their responses differently, you can get really, reaaaaaaaaaaally wacky results. For example, one single 19-year old black man's support of Trump in the USC/LA Times tracking poll moves the poll results for Trump up an entire percentage point on its own, and puts his support among black voters in the poll into the double digits (lol no), even though he's one of 3000 panelists (http://www.nytimes.com...).

What does this mean? It means, quite simply, that some polls are better than others. It is a terrible, terrible idea to take the results of any one single poll and hold it up as evidence of a Clinton landslide (Clinton +6 is one result in RCP from one of the most reliable recent polls with a huuuuuuuuge sample size http://www.realclearpolitics.com...) or a Trump triumph (Trump +6 in the USC/LA Times poll http://www.realclearpolitics.com...).

Looking at RCP is one way to gauge the polls--aggregates mean that polling errors on the whole have less effect. However, if you really want to have an accurate picture of the race, I highly recommend 538 (http://projects.fivethirtyeight.com...) as vastly superior to RCP. It actually takes into account the merits and biases of each individual poll, and adjusts/weighs them accordingly. I realize that some will not like this solely on the basis that Clinton has a slightly wider margin on 538 (3.4%) than on RCP (1.9%), but at the very least, just bear in mind that a poll by itself is pretty much nothing. Look at the whole enchilada (I have a thing for mixed metaphors now #sorrynotsorry).

So what you're saying here is that polls aren't reliable?

Polls are quite accurate. He is giving some good information, but is not saying that polling does not work.

I didn't mean that they don't work. I was saying "unreliable" in a sense of potential bias or strange accuracy. I meant it in a sense of "while the polls make it look like candidate A will win over candidate B, they are close enough that we can't confirm who will win"
You're probably thinking right now "haha I'm a genius". Well you're not -Valkrin

inferno: "I don't know, are you attracted to women?"
ButterCatX: "No, Vaarka is mine!"

All hail scum Vaarka, wielder of the bastard sword, smiter of nations, destroyer of spiders -VOT

"Vaarka, I've been thinking about this for a long time now," (pulls out small box made of macaroni) "W-will you be my noodle buddy?" -Kirigaya