very true This is going to be a pain in the ass. I'm on dial-up in the middle of nowhere for the next 7 days! vacations suck!
Some news regarding the impact the conventions have had. Too early to get a good read on the impact of the RNC, but the Gallup daily tracking poll shows Obama expanding his lead by a point. Will it change today? - Obama 49% - McCain 42%. Another item of interest, Gallup shows unhappy Hillary Clinton supporters deciding to swing to Obama, post-convention (Dem.), with 81% now, versus 70% pre-convention. That's old news, but worth repeating. Be interesting to see if that changes, post RNC. http://www.gallup.com/poll/110032/Gallup-Daily-Dent-Obama-Lead-Far.aspx Good news for Democrats, and dove-tailing with my prediction of a close popular vote, but an Electoral College landslide for Obama. On Electoral-vote.com, Obama is at 301 and McCain's at 224. That's up from yesterday. And folks, in my book, that would be a landslide. (cue Fleetwood Mac) http://www.electoral-vote.com/evp2008/Pres/Pngs/Sep05.png
How often is the Gallup Poll updated? Batman, don't get too confident right now. It's a long way to November.
Gallup and Rasmussen are updated daily. Rasmussen shows some movement in today's poll - from +5 to +2. That includes a day of polling after Palin's speech, but none of the polls today will have polling from McCain's speech. And since they are rolling averages, it'll be about Monday before both of these polls fully incorporate all polling after all the convention activity. With just one day, it's hard to say if it's statistical noise or something that will be sustained, but it's definitely a good sign for McCain. To get a 3 point move in Rasmussen (which is a 4-day tracking average), you need about a 12-pt differential between the 4-day old polling that was dropped and yesterday's polling. This ignores rounding - the actual differential could be as little as about 1.1 pt or more than 3 pts - and it's unknown whether the polling 4 days ago might have been really good or bad for one of the candidates due to random sampling. Give it until Monday and you'll have a good sense of where the race stands - though barring other factors, a McCain bounce will fade to some extent over the course of next week (bounces always get less pronounced over time).
Polls are a waste of time and effort because of several factors: 1) Political correctness -- Very few people will tell the truth for fear of being branded in some way. 2) Lifestyle changes -- Random cell telephones numbers can't be used because of federal laws on spamming and because the pollster can't tell if he/she is talking to a real voter or a 14-year-old who claims to be. Remember, a growing number of households don't even bother with landlines anymore. 3) True indepedent polling -- Most pollsters are eager to provide the positive results that their clients want to hear. Afterall, it's for propaganda purposes.
None of this is true. The polling firms have reputations to uphold - they have a very clear financial incentive to provide as accurate polling as possible. Many also publish their methodology for scientific review. It's not like they provide data that won't be proven - there will be an official accounting of their results (voting), so they have a very real incentive to be as accurate as they can possibly be. This whole "pollsters are biased" thing is nonsense promulgated by people who don't like the results of a poll (not you, but when its pushed by campaign officials and the like). Now, there ARE biases inherent in various polls - due to their methodology and the like. But even that can be analyzed. Pollster.com does exactly this: http://www.pollster.com/blogs/how_pollsters_affect_poll_resu.php It includes both the numbers as well as the reasons why it happens. The two tracking polls above for example, have 2 of the strongest McCain bias - it's not due to trying to be biased, but just how it plays out.
That's why I said "most" (not all) pollsters. Some of the big ones have reputations that are solid, I agree. Unfortunately, not all pollsters are like that. Also, you said, "None of this is true." How are reasons #1 and #2 untrue?
I'll chip in my 2 cents about #2. I refuse to have a landline. We had one for a few months, but the only calls we received on it were people asking for donations. We quickly realized how useless the landline was for us, considering we had cellphones.
the republicans are def. doing a great job POLITICALLY with their moves. this is going to be as tight of a race as 2000. if the democrats can't win this yr, they just need to put hillary out there. palin is becoming a rock star (and she has to thank the media for that). but man, after the new unemployment reports today, and mccain wins, i guess america IS happy after all.
Didn't mean to click submit :/ But, yeah, I definitely see a demographic being underrepresented in polls because of the cell phone/landline. as far as #1? Sadly, I agree, hopefully this issue will change some day (maybe in my lifetime.)
My point exactly. Fewer and fewer of us have landlines. I have a VOIP telephone so I can call all over the U.S. and Canada at no charge, but, if I didn't need that, I too would be cell all the way. However, neither party will be calling you for your political opinion.
On #1, many of the polls are done using automated telephone systems to avoid the human bias. And in the primaries, there was no systematic under or overestimating of Obama votes for the polls. On #2, most of the poll companies do demographic re-balancing. They account for past trends and the idea that they won't reach a sampling of people that matches the actual population. For example, if they random sample 1000 people and get a whole bunch of elderly or black or Democratic or religious voters, they'll adjust the results to better reflect the real demographics of the region they are polling. It's a very scientific process - yes, it has problems, but to say they are worthless because they can't be exactly perfect is silly, in my opinion. That said, I think their biggest value this early in the process is to see trends. That accounts for all the error created by methodology. Poll A has the same methodology when taken last week vs this week - so if it shows a 5% gain for a candidate, that's useful information regardless of the starting point of each.
We can argue this forever. However, let's say I'm a racist opposed to having a black or brown person in the White House regardless. Do you think I'm going to give my real opinion regardless of whether the call is automated or human? Next, have you ever worked in a Democratic or Republican call center during election time? Let me give you a clue -- no cell telephones. If it is not in the land line telephone directory, that number is not going to be called. That alone skews results toward older people rather than the culturally hip.
There are scientific studies that account for cell phone use, and random digit dialing studies that include cell phones. I am sure the top polling firms are trying to adjust for cell phone use if not directly calling cells phones. The biggest challenge IMO is accounting for refusals (nearly 1 and 3 may decline). You can also try to get adjustments, try to figure out if any group is systematically refusing more than others, but it is tough. A big one issue to me that doesn't get much play from what I can tell is how do they determine "likely voters". Are likely votes really that much more likely to vote than registered voters, are likely voters overconfident that they will get to the polls and those not considered likely voters under-reporting their future voting? Who really knows but my sense if polling firms have not paid enough attention to this. So yeah there are biases involved in any methodology. I think the best you can do is a poll of polls that include LV and RV samples, and probably weeding out the extremes or those with particularly bad methods might make it even better. Not perfect (I don't think scrutinzing the method or taking out extremes) , but I do like the poll data from this cite for a quick estimate...http://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html I think we need to wait until about next Thursday or Friday to get a solid reading on where we are at. This will account for media play this weekend post convention, and adequate time to process the conventions (so all the rolling poll data will be next monday on----).
Again, we have actual polling data that correlates to actual results from primary elections - and it doesn't show any systematic problem of this sort. Besides which, yes, there IS a good chance you will give your real opinion. Something like 20% of Democratic voters in Ohio actually told exit pollsters that race was an important factor in their picks - and that was to a live human being in person. And again, the pollsters KNOW this and account for that in their results.
By the way, for those of you interested, 538 tries to back out the daily results from the Gallup/Rasmussen daily polling: http://www.fivethirtyeight.com/2008/09/crackin-trackin-poll.html (Lots of charts, so I won't try to copy the article) Notice just how volatile the individual day polling is. In a 3-day period, they estimate it was Obama +11, McCain +2, Obama +6.
This is the key to the errors in the primaries - the turnout was all whacked out with young voters voting in record numbers, etc. The other thing is, even if they divide Likely and Unlikely voter, they can't measure enthusiasm or ground game all that well. Up until this past week, both of those substantially favored Obama. I think Palin, at least in the short term, has neutralized the enthusiasm issue though. So I think if an election were held last week and this week, the results would be very different even if the polls were not.
I'm guessing if you're a racist, you're not going to lie to a machine and say you're voting for Obama. Why couldn't you just say you're voting for McCain?
Good point -- I was thinking exit polls. On the telephone there are a lot of people like me who just like to mess with the pollsters and there are those who don't want to give opinions for fear of getting bombarded with requests for contributions based on recorded / non-recorded responses.
It appears Palin did indeed cause a Republican bounce as Obama's lead is down to 4 points. Seems like she influenced undecideds. Since this only includes one day of polling after Palin's speech, it seems likely that the lead will continue to decline and we are once again statistically tied.
