Skewed results in the polls?

>> our deep dive is the most famous name in polling giving the president a raw deal? fairly or not, when gallup speaks, it can steer the conversation and have outside impact on the conventional wisdom of the presidential race . ever since gallup competed with the robo pollsters by launching the daily tracking poll, including myself, have raised serious concerns with the methodology. they say it has shown the president under performing compared to other polls. every poll, slight variations matter. even how you word a question and the question order. they posted a review of gallup polls of the president over the past year. what it found was that gallup found obama with a lower job approval rating than that of the other top organizations using similar methodologies. he averaged among the top six, 47%. the average with gallup polls, 44.4%. the pollsters said it's because they measure responses from nonwhites differently than other polling organizations. the method of tweaking numbers to account for demographics and account for various groups that make up the country as a whole. here's how it works. say you are polling an area where men and women is split 50-50. 3/4 of your respo women on your dave calling. in calculating the results, you would boost the importance of the men's answers and reduce the women's answers to reflect the 50-50 split. they do it for a number of factors, but where we are concerned it's about race. according tote survey, it doesn't matchup with the latest census numbers from the government. for instance over three months, gallup 's polling data was weighted to represent a population that is 11.3% african-american, short of the census figures that show the break down is 12.1%. the polls reflected hispanic population up 12.4%. why is this significant? these are the types of voters that support the president in big numbers . gallup found in may he poled at 54% among hispanics which is low, but 90% among african- americans . taking any of the voters out of the mix can change to the perception of the race. right now i'm joined by the two men best equipped to discuss the topic. ed newport and mark bloomenthal, founder of pollster.com and the man behind that analysis that we did. frank, my friend, let me start with you. what do you say to the criticism and critique of your methodology?

>> i'm glad you are talking about this. it's a complex cross. you need about an hour to explain it.

>> we use a special report . it's good that we're all looking at it. it is so complex that we think race is not the only variable involved. when we are rating polls, the most complex thing is cell phones and we plan to intrusion 400 up to every 1,000 based on cell phones . we have to wait age and education and gender. all of these are correlated with presidential views. political views as well as race. the " huffington post " seems to be monor focused on race, but for our methodologists and statusticians to adjust, we have to take into account more than one number. actually they're based on over 100,000 interviews. for blacks and hispanics, you have to put it into the mix and avoid measurement era and designed effects. we can wait blacks up to be 30% of the population and our methodologists make the decision that when we balance all the factors out, we end up with what we think is a solid sample of the american public.

>> okay. i understand that, but race he said is it sounds like it was too much of a factor. i have to say in these numbers, it seems to be the way that we polarize and saw the analysis and race is dividing on political lines in a way we have not seen for a long time.

>> it is one of many factors that a pollster needs to adjust. this is a problem facing not just gallup , but all polls. the low response rates means that they have a harder time finding voters and americans in urban areas that means it's harder to find not only nonwhites, but younger and less well-educated. the way we weighed those is important to getting accurate results. race i think is the most important of those because as you showed, the biggest differences in political attitudes come when you slice americans by racial and hispanic groups.

>> when you look at the exit polls , it's a simple thing. you talk to the two campaigns and it's simple for them. if it's at one point 78% of your registered voter population as white which it hadn't been for a decade in some of your polling about a month ago with 76% i believe last time and expectation of 74 and 73, that can alter the numbers in a giant way. can it not?

>> yeah, race is clearly an important factor. i don't think on average we have the white population that high. no question that race is an important factor. so are the other variables. age is highly related to the context as well. so is education and gender and so is having a cell phone only or land line . all of these are important factors. our statusticians and methodologists have been doing this for a long time and take those into account and try to put our samples and give the best overall sample that we can come up with based on as i agree with what mark just said. when we bisqually do our introduce, we under represent certain groups and have to give that when we rate the data.

>> do wish you were not daily tracking?

>> the major contribution we are making here which i think is important is large sample sizes. it is based on 3,000 interviews with registered voters and we provide analysis based on 9,000 interviews based on three-week rolling average . we are able to give people a stable picture of what's happening with the race as opposed to other polls where we saw obama was ahead by 13 points. clearly an out liar and daily tracking with large sample sizes, we think that we are making a real contribution to understand the trajectory, the trend line of the race.

>> it's funny you bring up the out liar and bloomberg. we had criticism at that. she responded to it. one of the criticisms we talked about, the biggest difference in that poll and why obama had a bigger lead, again it was the race break down. she only had the white population under 70%. which we hadn't seen either.

>> one credit to other polls including nbc and " wall street journal ," at the end of the survey, every time you release the results, you can see all the demographics. you can see what the pollster hit the targets and gallup is fabulous in disclosing all sorts of things, but you can't see for those 3,000 interviews every day and six or 7,000 a week what they were weighted to.

>> two other things and i will run out of time and my producer will be mad at me. let me ask you when it comes to waiting, do you wish you could wait by party?

>> no, we don't. that's an old thing. you may remember the good old day when is that used to be the topic du jour when i talk with people like yourself. that's a moving target . a lot of them say i know what they will be like on election day . i'm not sure they do. the way they manage the party may be different than the way we do it. we stayed away from the effort to try to wait to an unknown moving target .

>> let me ask you this. i have been an advocate and i can't talk them into doing it and talk them into doing this. we have discussed this. why not release multiple results. this is what your results would look like with the 2008 exit poll and the 2010 . what we think it could be. why not release three versions every day?

>> that's an excellent idea. you are right. what you are talking about is scenarios. what if this and what if that. that's an excellent way to do it. we may do more of it this year. thanks for the suggestion.

>>> we confusing the situation?

>> we do a good job confusing people either way . it's important to see both if the result is different, it helps for us to be able to look at the poll and see what the democrat-republican mix was and the racial mix.

>> transparency, transparency. that is one thing that you guys are helpful at giving everything out. no one can say that. mr. newport, a pleasure as always. thank you. i'm sure we will be talking more