question regarding poll fluctuations

I'm curious if anyone who knows more about polling and psychology than I can shed any light on a question I have.  Why is it that poll numbers fluctuate so much from poll to poll?  More specifically, why is it that people are so willing to change their minds from one poll to the next based on one specific event?  I'm specifically thinking here of the small uptick that Bush seems to have had following the Zarqawi death?  We always seem to see these sorts of upticks following specific events like this.  My question is why?

I'm the sort of person that pretty much doesn't change my mind from day to day on whether I think a politico is a jackass.  If I think someone sucks bad enough to tell a pollster I don't approve of them, I'm sure as hell not going to change my mind because of one little thing to the point where I love them the next day.

I realize that the same people are not being polled during each poll.  I also realize that most people are pretty casual news / politics followers.  However, these shifting poll numbers clearly show that opinions change based on events.  How and why does this happen?

Anybody have any light to shed on this issue?

•  I think it's an easy answer...

  for whatever reason, most American's have this need to WIN.  To be on the WINNING side.

  When they see the wind blowing the other way, they don't want to be downwind from the smelly losers.

  by kishik on Fri Jun 16, 2006 at 03:55:19 PM PDT

  •  Sortof like Afghan Warlords? {N/T}

    by grndrush on Fri Jun 16, 2006 at 06:38:49 PM PDT

    •  sort of like in many places...

      but definitely prevalent in US culture.

      by kishik on Sat Jun 17, 2006 at 03:06:56 AM PDT

•  Lots of reasons

  Margain of error.  A good poll had a 95% chance to be within the MOE.  But even within it, there is a big range.  Say the results 37 for Bush.  He is really somewhere between 34 and 40.  In Bush's case, this could mean he is getting no uptick, or that he was not so low as his earlier polls showed, or neither one.  The current change is too small.

  No two polls ask questions quite the same way, so comparing one company's poll to another's can be a problem.  

  When was poll taken.  Day of the week and time of day shifts who responds.  

  Big changes can be due to people's emotions basd on recent events.  People do get psyched up or depressed based on a lot of thinfgs and it shifts results a bit.  

  Advice:  Ignore the small changes and just look at the long term trends.   For example, today's WSJ poll shows Bush has improved from 36 to 37.  But his numbers starting last September are:  40; 39; 38; 39; 39; 37; 36; 37.  Realistically, considering the MOE alone, there has been no change.  He is stuck between 35 and 40.

  by hilltopper on Fri Jun 16, 2006 at 04:19:10 PM PDT

  •  We have a winner!

    Margin of error is the correct answer, but for a funny spin read this comment. The last paragraph is priceless.

    by TerraByte on Fri Jun 16, 2006 at 06:43:52 PM PDT

•  I think

  it's just that the most recent events are the ones that most affect us.  You've got some fraction - say a third - of the people who are resolutely on one side of things, another third who are resolutely on the other, and the remainder floating around somewhere in between.  The first two groups aren't going to waver much - maybe a little between Agree and Strongly Agree, but otherwise pretty much stationary - but the third group isn't particularly happy on either side.  The most honest answer most of the time is the one the pollsters most hate - and take pains to avoid - which is "I don't know or care."  If they're pushable (and since they're not committed to either side, I think this group's most likely to be so), they'll slide away from center - and the event that they're most aware of is whatever just hit the newspaper.  

  Probably needless to add: this is all uninformed speculation.  Ask me tomorrow and I'm sure I'll have a different answer. ;-)

  by cmanaster on Fri Jun 16, 2006 at 04:26:30 PM PDT

•  right, it's in the reaction

  One week the people answering (and 1 in 10 here is what we're talking it about) say to themselves 'I think things are going badly in Iraq' and answer one way; the next week, they may think 'Well, we did get Zarqawi,' but they may not have changed their long term thinking at all.  Remember that answers to questions are incredibly sensitive to how they are posed.  Presumably it is just like asking the question with a positive comment attached to the phrasing.

  by John DE on Fri Jun 16, 2006 at 04:30:18 PM PDT

•  Hey! It's good news.

  If they wiped out a small time thug turned terroist, it is good news.  This slug has killed dozens of innocent people.  Killing him was a good thing. Sure Bush gets an uptick for this.

  by biotecchie on Fri Jun 16, 2006 at 04:33:00 PM PDT

•  It's just statistics

  All (reputable) polls come with a margin of error attached. What this means is that there's likely some spread between the number that the poll reports and the actual number if you went and asked everybody.

  So, if you see a poll today at 35%, and one next week at 37%, and both have margains of error of 3%, those polls are statistically identical.

  To make any real sense of polls, you have to look at things like long-term trends, rather than quick comparisons between nearly-equal polling results.

  -dms

  by dmsilev on Fri Jun 16, 2006 at 05:12:15 PM PDT