HomeLatest ThreadsGreatest ThreadsForums & GroupsMy SubscriptionsMy Posts
DU Home » Latest Threads » Forums & Groups » Main » Politics 2014 (Forum) » I Have A Question Re: Pol...
Introducing Discussionist: A new forum by the creators of DU

Sat Oct 6, 2012, 05:53 PM

I Have A Question Re: Polling And Methodolgy

I saw a post by a fellow DU member that suggested Romney got a twelve point bounce in the Gallup. Rasmussen, and Ipsos Reid tracking poll. And then another post suggesting the same thing.. At first I couldn't imagine where those numbers came from. However it seems the person looked into the length of the tracking poll and inferred from that single day movements.


Is that even possible?

6 replies, 673 views

Reply to this thread

Back to top Alert abuse

Always highlight: 10 newest replies | Replies posted after I mark a forum
Replies to this discussion thread
Arrow 6 replies Author Time Post
Reply I Have A Question Re: Polling And Methodolgy (Original post)
DemocratSinceBirth Oct 2012 OP
TroyD Oct 2012 #1
LiberalFighter Oct 2012 #2
DemocratSinceBirth Oct 2012 #5
Cicada Oct 2012 #3
titaniumsalute Oct 2012 #4
FrenchieCat Oct 2012 #6

Response to DemocratSinceBirth (Original post)

Sat Oct 6, 2012, 06:55 PM

1. Bumping thread n/t

Hope someone can answer your question.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to DemocratSinceBirth (Original post)

Sat Oct 6, 2012, 06:59 PM

2. Did the poster even provide a link to the data?

I checked Gallup and did not see a 12 point swing.

IMO there is no way there would be that big a swing in that time period.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to LiberalFighter (Reply #2)

Sat Oct 6, 2012, 07:50 PM

5. It Has Something To Do With Days That Are Added That Are Dropped Off

But I don't think it's possible.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to DemocratSinceBirth (Original post)

Sat Oct 6, 2012, 06:59 PM

3. I don't think it's possible

If every day in the tracking period were identical then you could figure out the one day results when the spread changed. But consider this example - a three day period is used for the reported spread, with the oldest day replaced by the newest day each new day. What if the oldest day had Obama leading by 20, the middle day had them tied, and the most recent day had Romney leading by 20. The tracking poll would show them tied. Then there was a debate and on the day after that Obame led by 10. The updated tracking poll averaging the 3 most recent days would report Romey leading by 3, a 3 point gain, even though Obama did much better the day after the debate.

So unless you know the individual days results I don't think you can determine the most recent day's results from a change in the three-day average.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to DemocratSinceBirth (Original post)

Sat Oct 6, 2012, 07:10 PM

4. Here's some info

Gallup Daily Tracking...

They use a 7 day rolling average. SInce the Sep.t 19-25th survey result there have been 10 results. Each day one day drops off, another is added on. They do this to stabilize the data by having significant sample sizes.

He has fluctuated between 49% and 50% each day of those 10 results. So there's no way he could have lost 12 points. 2 days of sample are now post-debate. Romney has had 3 days at 45% and now 46%. All I see is some tightening in the Gallup poll.

Ipsos. They use a 4-5 day tracking poll average again dropping a day and adding a new one each day. Since Monday, Obama was at 46% and today he went to 47%. (two days of that poll are post-debate.) Again, Romney's numbers have come up some and it has tightened. But there is not going to be a huge swing in the next day or so.

Rasmussen. It is a piece of shit poll. They DO NOT call cell-only households which skew younger, more ethnic, lower income. Those demos are typically more Democrats. In addition, they use automated interviews by computers. People can easily screw with the results, lie, etc. There is no randomness to their sample which is a problem. Kids can answer the phone and push the buttons.

I worked in survey research for 10 years. I know how this stuff works. It is too bad that people spend lots of time opining about shit they know nothing about. It tends to whip people into frenzies many times for no reason.

Reply to this post

Back to top Alert abuse Link here Permalink


Response to DemocratSinceBirth (Original post)

Sat Oct 6, 2012, 10:18 PM

6. I simply would say that vigilance should be upmost to make sure we don't buy BS....

Results and Interpretation of a Study of Effects of Poll Reports on Voter Preferences: The Mehrabian Polling Snowball Effect
Mehrabian, A. (1998). Effects of poll reports on voter preferences. Journal of Applied Social Psychology, 28, 2119-2130.

Overall, results obtained from both studies were consistent in showing the superior strength of the bandwagon or rally-around-the-winner effect .... it is important to consider cumulative effects of the bandwagon effect when it is combined with repeated and closely spaced reports of polling data. Assuming that the bandwagon effect is operative and one candidate is an initial favorite by a slim margin, reports of polls showing that candidate as the leader in the race will increase his or her favorable margin. Subsequent reports, based on more recent and stronger margins, will in turn progressively strengthen that candidate's lead.

This unstable equilibrium effect of polling described in the preceding paragraph, or The Mehrabian Polling Snowball Effect (MPSE) , helps identify a potential way in which political organizations can be tempted to influence voting by sponsoring biased "polling studies" and reports. Poll results can be slanted easily through selection of slightly skewed respondent samples or the actual wording of questions used in the polls. Frequent reporting of slanted and invalid poll results can help propel a candidate to the forefront and, in fact, increase his/her lead over time. Similar considerations would apply to major political issues (rather than candidates) is various political campaigns.

Accordingly voters need to be educated about the Polling Snowball Effect so they can be specially vigilant when they are repeatedly barraged by polling reports favoring one candidate (or poll results that suggest popularity of a particular campaign issue) during political campaigns. Voters, in particular, need to educate themselves about the political orientations of entities that repeatedly sponsor polling studies.

http://www.kaaj.com/psych/abstract/pollsabstract.html

Reply to this post

Back to top Alert abuse Link here Permalink

Reply to this thread