Page 7 New CNN poll ..NOBODY POLLED UNDER 50yrs Old

[Ichingcarpenter]

Get that? ........... nobody was polled under 50 years old


http://i2.cdn.turner.com/cnn/2015/images/09/20/dempoll.pdf

[Erich Bloodaxe BSN]

1. That's Hillarious!

So in order to maintain the 42%-24% lead, they had to sample exclusively from people with whom a certain candidate does better, and exclude those who are tending to support a different candidate?

Bwahahahah.

[rpannier]

54. You mean

that's Hillary-ous as in do whatever it takes to make her look unbeatable

[peacebird]

81. Absolutely Hillary-ous!

[Snotcicles]

95. Looks to me like non were under 65. Edited

That was for second choice.

[merrily]

129. Over 65 AND didn't really care all that much about expanding OASDI, I assume.

[still_one]

119. It was of REGISTERED Democrats, and what it means is that Bernie supporters had better register to

vote, because you cannot vote unless you are registered.

So laugh away

[peacebird]

124. We laugh as we register new voters, we laugh as we get voters to sign petitions to get Bernie on

Ballots.

We Sanders supporters laugh a lot!

[still_one]

131. Well, this poll indicates that there were NOT a significant number of REGISTERED Bernie supporters

You don't think it is valid, that is your right

[PotatoChip]

206. Or it could just simply mean that the polling firm

was instructed to ask for registered voters aged 50 or older. If no one within the household fit the criteria, the pollster would have politely disengaged and moved on.

Different demographics of all sorts, including age, are often selected for (whatever) reason.

[still_one]

207. Possible, or it can indicate that people under 50 are not registered to vote in meaningful numbers

yet according to the poll.

That could be a possibility, since the midterm turnout was 41% the turnout in 2010, and the lowest in 70 years.

My only point is that it would be in everyone's interest to get as many people registered to vote, especially in the less than 50 year old demographic

[roguevalley]

218. I've been dumped over age and other criteria. They polled for what they wanted

and weeded out the inconvenient truth.

[Gormy Cuss]

276. No. You were screened out because they had already interviewed enough people like you.

Had they just polled without screening for gender,age, race, or other key demographics the poll results would be heavily skewed. For example, women are more likely to agree to be interviewed than men, and seniors more likely that twenty-somethings. If the polls are conducted with the first 1000 people who agreed, the researchers may find that they have a mostly female, mostly older set of opinions, which limits the usefulness of the data. By using screening questions up front the pollsters can control for that upfront by actively seeking out people who are from other demographics and in the end have a nationally representative sample.

[840high]

256. ...^ that

[PatrynXX]

258. that wasn't well written was it

yeah I can laugh at how after Obama set records registering young voters they suddenly aged 30 years. this isn't a Soap Opera...

[merrily]

128. Then tell DWS she should have scheduled the first debate earlier than she did.

[still_one]

134. That is a different issue. The respective campaigns should be working to register as many voters as

they can

Even if it means sending out a cost effective email to their supporters to get them to push other supporters to register

[merrily]

141. Very related issues. She's depressing the vote, esp among Bernie's demographics.

I am sure the campaigns are doing what they can to registers voters. I do that for every election, no matter what. I work on getting absentee ballots for those who need them, too.

[still_one]

144. No offense, but if that is the case it sure represents a shallowness and lack of maturity in the

potential voting public, and they will get the government they deserve

[merrily]

155. Why would your incorrect statement about the voting public offend me?

Everyone is not a politics addict. People have day jobs, sometimes several, families, hobbies other than politics, etc. This is no surprise to anyone. This rigging the debate schedule shit is on DWS.

[sabrina 1]

245. No offense, but when the chair of the DNC suppresses the people's right to hear all candidates

I don't think it's the voters who are immature NOR do they 'deserve' to be bamboozled by someone who has already lost us the House and Senate and is about to lose the Dem Party, the WH. This is pure corruption, so blatant that anyone can see it, and they ARE seeing, by the millions. And one thing is for sure, it isn't helping her candidate.

[questionseverything]

165. not all registered dems..from the poll

BASED ON 261 REGISTERED VOTERS WHO DESCRIBE THEMSELVES AS DEMOCRATS AND
131 REGISTERED VOTERS WHO DESCRIBED THEMSELVES AS INDEPENDENTS WHO LEAN
DEMOCRATIC, FOR A TOTAL OF 392 REGISTERED DEMOCRATS-- SAMPLING ERROR: +/- 5%
PTS.

[still_one]

171. True. The poll indicates that there are not enough registered voters under 50. I read that as

there needs to be an effort to register Democrats for voting

[questionseverything]

174. dems should always be registering peops to vote ,that much is true

but the pollsters know the age demographic before they call

the under 50 set was definitely under sampled

[Aerows]

237. Well considering that in quite a few states

that deadline to register is before we even have the first Democratic debate, which is where a lot of voters get their first view of what the candidates stand for, you can thank the idiots that planned the schedule if it's hard to get people registered.

How in the hell does the DNC think it is benefiting GOTV efforts if people haven't even seen the candidate before the registration deadline?

Or has everyone decided that Democrats have a better chance to win if fewer people are registered to vote?

[OilemFirchen]

241. Which states require registration more than a year before an election?

You don't have to list them all. Just a few would suffice.

TIA.

[Aerows]

242. To vote in the Democratic Primary.

New York - deadline is in 17 days.

Source: http://voteforbernie.org/

[OilemFirchen]

249. Your post said nothing about Primaries.

Nonetheless, one is not "quite a few".

[Aerows]

251. Oh well, OilemFirchen

You caught me. I didn't mention the word "Primary" and I used the word "few".

Please feel free to give me plenty of lashes with a wet noodle.

How dare anyone complain about the debate schedule.

That's perilously close to criticizing Hillary Clinton and good Dems should shut up prepare to vote for her.

[Robbins]

2. MSM cooking poll

they didn't poll Bernie's best group In order to have headline Hilary lead grows.

the Hilary supporters on DU won't mention it.granted i put a lot of them on my ignore list.

[merrily]

132. So many things about this season have been rigged against Bernie, it's disgusting.

[Snotcicles]

153. I gotta a survey/begging this weekend from the DNC so I took

my extra bold sharpie and wrote "REMOVE ME FROM THE DNC's MAILING LIST (I'll reconsider if the DNC becomes Democratic again"

[merrily]

162. Great, when you can do that on something asking for money. Or a call.

They're getting like crooked televangelists who took money and checks out of envelopes, then discarded the envelopes with the prayer requests still in them. What DWS did with the debates proves they could care less. So do all the bs memes they try to use to brainwash.

Save your money for individual candidates---although they've rigged so even donating to an individual candidate gets money to DINOs in Congress.

[Snotcicles]

164. Yep you know they are going to look at it because they want credit card #'s. nt

[TexasTowelie]

3. Also interesting is that everyone attended college.

Talk about stratified samples.

[abelenkpe]

4. Nobody under 50 would answer a number

From a source they don't recognize. Telephone polls will always skew older.

[Ichingcarpenter]

6. You know that because?

Of what data ?

[99th_Monkey]

46. Probably only polled land-lines, I'm guessing, not cell phone users. nt

[Honeylies]

66. Cell phones were included

"This sample includes 606 interviews among landline respondents and 400 interviews among cell phone respondents." See page 1.

I believe most top political polls since 2008 have included cell and landlines.

Anyway, awesome catch by OP on the age skew for this poll. I wonder how they managed to exclude younger participants? I can't imagine it was luck of the draw, did they target them?

[99th_Monkey]

68. Cool. I guessed wrong. my bad. nt

[roguevalley]

76. I have been asked age and dropped because I am

Too old. They do age skewed polls. So sad, this

[demwing]

67. Guess again...read page 5, first paragraph

"A total of 1,006 adults were interviewed by telephone nationwide by live interviewers calling both landline and cell phones"

http://i2.cdn.turner.com/cnn/2015/images/09/20/dempoll.pdf

[99th_Monkey]

69. OK. Thanks for clearing that up. nt

[JHB]

80. About 40% were cell phones, according the linked PDF

Interviews with 1,006 adult Americans conducted by telephone by ORC
International on September 17 - 19, 2015. The margin of sampling
error for results based on the total sample is plus or minus 3 percentage
points.
The sample also includes 924 interviews among registered voters (plus
or minus 3 percentage points).
This sample includes 606 interviews among landline respondents and
400 interviews among cell phone respondents.

However, the "N/A"s for the age groups under 50 (and some other breakdown categories) are more than a little curious.

[99th_Monkey]

86. Got it. Thanks. nt

[passiveporcupine]

238. Should I tell you one more time

Just in case?

[Erich Bloodaxe BSN]

49. I will, but I haven't lately.

On the other hand, I'm 46 or 47 (I don't bother to keep track much these days) so I'm closing in on 50, and I do have a landline. I like answering polls. I did my part a few years back to claim curling and archery were the most popular winter and summer olympic games among left leaning midwestern follks

[Mr.Bill]

99. When election day comes

people under 50 won't vote when they find out they can't do it on Facebook.

[IronLionZion]

244. Personal experience

I've gotten numerous missed calls from Gallup who for some reason decides NOT to leave a voicemail or call back number to let me take the survey when I come home after work. So instead, I managed to take a cell phone survey while driving home one night, against my better judgment. It's an idiotic practice that definitely misses a lot of people who work during typical working/commuting hours. So those types of surveys would skew towards older or stay-at-home types.

[SonderWoman]

5. And they included Biden.

It would be nice if they waited until Biden announced.

[MADem]

7. Some one w/polling expertise needs to translate this.

I think they interviewed them, they just can't be sure of the accuracy of the results with them...?

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of
+/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with
an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error
larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A"

I don't know if this poll is very accurate, in any event, with such a big sampling error.

[Ichingcarpenter]

8. I took two semesters on polling for political science

the poll is worthless even their sample size is marginal never mind their chronological absence of a real sample of the the voting public.

[MADem]

18. It looks like they did interview people of all ages, not sure where the data is though.

According to this poll --assuming we take N/A as meaning "not included" (and I don't think that' s what is meant, here) --they only interviewed people with college educations, and they aren't from any segment of the country either, if we're looking at all those N/As.

I think they're using all the data but they're only putting their money where their polling is on some segments of it.

I don't think polling data is worth an awful lot at these early stages--they pretty much show who's getting talked about...or more significantly, who is NOT getting talked about. Webb and Chaffee need to go home if they don't start making the rounds soon.

[Ichingcarpenter]

29. N/A means not available or not applicable

The poll is worthless

[virtualobserver]

45. good catch- 18-50 is half of the US adult population

[thesquanderer]

50. That's not exactly what N/A means in this case

Again...

results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A"

So it doesn't mean that there was no info on those groups, or that that info wasn't used in the aggregate figures... only that the statistical accuracy for that subgroup is not sufficient for them to provide the number broken out by itself.

[demwing]

70. the #1 reason for that high of an error margin

is because the sample size was too small.

In other words:

"We didn't poll enough people under 50, therefore the MOE was extremely high. So, instead of polling more under 50 voters, we just ignored them."

Shameless...

[DallasNE]

148. Great Point On "The Aggregate"

But some of the areas marked N/A are areas that demonstrate a serious problem with the methodology used. The 18-50 age group would be well over half of the voting age population as would those without a college degree so how could their sample size be too small to show a breakout. Does this mean that too many of these people refused to be polled or that they knew the demographics ahead of time for those called. Whatever the reason, this poll is highly unreliable because of the lack of balance in those polled.

[thesquanderer]

289. I debunked it first ;-)

Really, it's self-evident to anyone who actually reads the poll.

[Gormy Cuss]

275. In this situation it means that reporting the data sliced that finely would be irresponsible

because the margin of error is too large to report the opinions as nationally representative.

It does not mean that the overall poll is worthless. It means that the methodology was constructed to provide info on opinions at broad level but not at finer cuts. It would be easy to do the latter, it just costs a lot more money to construct the sample and to find sufficient respondents in each targeted category (i.e., quota group) to minimize the margin of error.

[jeff47]

74. My guess is they got so few respondents that they couldn't get

a statistically significant sample size.

So they got a few non-college-educated, and a few younger-than-50, but not enough for their answers to be usable. So they weren't used.

[thesquanderer]

133. They were used.

Again looking at the text supplied, it sounds like people from all categories were used in determining the overall results; but in some cases, there were not enough people in a particular demographic to be able to make a meaningful statement about their preferences apart from the whole.

[jeff47]

139. And they shouldn't have been without a bigger sample.

The results were crap, so they did not put them in a crosstab.

If the results are too crappy for a crosstab, they're too crappy to put in the "headline" results.

[zeemike]

61. Not worthless to some.

I imagine to some it is worth what it cost to get it made and more.
Popularity like consent can be manufactured.

[Ed Suspicious]

107. Yes.

[Bubzer]

259. +1

[stevenleser]

233. 1006 were asked about Obama. Only 392 Democrats & Leaners were polled about Democratic nomination

Breaking those down into age groups and various other categories would produce in some cases numbers that were too small to present as representative.

[MADem]

236. But that doesn't mean they didn't interview anyone UNDER fifty yes?

It's just that they can't get declarative about what direction those groups are trending towards...?

[stevenleser]

239. Correct. And we can make some assumptions based on...

"All respondents were asked questions concerning basic demographics, and the entire sample was weighted to reflect national
Census figures for gender, race, age, education, region of country, and telephone usage"

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of
+/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with
an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error
larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

and

http://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf <--- national census figures for gender and sex at least

which tells us that
22% of the population is age 18-34 that would have been only 86 folks out of the 392.

and

23% of population is 35-49 which would have been 90 out of the 392.

[questionseverything]

254. so 176 when you combine the 2 n/a groups or 45%

but looks more like 98 peops for both groups combined

the breakdown is 25% under 50,75% over 50

[stevenleser]

264. Not if they went with census figures as they say they did. It's 176. nt

[questionseverything]

268. we know the numbers (approximately) for 2 columns

8moe 156
8.5moe 138
=294 over 50

392-294=98

98 =25% of 392 total voters

of those under 50 voters, 47 vote for bernie ,37 hc,14 others

i was mistaken earlier thinking it showed even lower but a 75/25 split is not a good sample

http://www.lifestylesurvey.org.uk/05_methods/sampsize.html

[stevenleser]

269. I think you are trying to back into the sample size in a way that may not be right.

The only definitive thing we know about the sample size is that they weighed the sample to produce demographics that match the latest census.

[questionseverything]

270. are you saying the link i posted has wrong cut offs for moe?

or that we do not know the moe rates for 2 columns?

or that cnn mislabeled their columns?

or that i have confused which columns are labeled?

it is really simple math, if you think i am wrong explain why

[stevenleser]

271. Mine is also simple math and has the advantage of going by what they said they did.

Prove mine wrong first, then I'll work on yours.

[questionseverything]

272. for anyone that might be interested

pg 7 of the poll

look at the middle graph

the last 2 columns are under 50 n/a, over 59 +/-6.0 moe

+/-6.0 cut off is 277 sample size ,when the poll breaks apart over 50s into 2 groups we get an +/-8.0 and +/-8.5 with respective cut offs at 156 and 138,which is were i get the 294 number,which is with in the +/-6.0 grouping

we know the entire group polled consisted of 392 dem/dem leaning voters

392-294=98

98 peops in sample size would be a +/-10moe past the cut off cnn reports which is why there is an n/a label on the under 50 group

[Gormy Cuss]

274. It means they are using a best practice in terms of reporting results.

It's hard for me to describe this without jargon so I'll use a very simplistic and not totally accurate way to describe it. The poll has a relatively small number of respondents but enough to make reasonable national estimates of opinion by applying weights to adjust for small discrepancies in the number of respondents by category. However, once the data is segmented into subcategories (like a cross tabulation of support for each candidate by age group of the respondent), the actual number of people polled may be too low to report as representative of that group. Best practice is to state as ORC did that the sampling error was too large to report results. It does not mean that they did not interview people under age 50. If the poll had more responses, say 1000, there would be a better chance that cross tabs could be weighted to be representative.

[MADem]

279. Thank you!! That was a great explanation, it's what I

thought was going on, but I didn't have the verbal skills to say what I was thinking!

That was a very clear explanation--if you don't teach this stuff, you should!!

[DetlefK]

9. An overview on who was polled:

These are the people who were polled about the democratic primary:
- nobody under 50 (page 7)
- nobody without college-degree (page 7)
- nobody from rural areas (page 8)
- no non-whites (page 9, but on page 11)
- no Independents (page 10, but on page 12)

[SonderWoman]

14. Shouldn't that make this poll skewed towards Bernie?

White, city liberals, with college education.

[TM99]

17. Disproves that mythic meme, doesn't it.

[restorefreedom]

19. millenials are for bernie

this skews it to hillary

[SonderWoman]

31. Millenials are split about 50/50 and...

Are also the least reliable voting bloc.

[restorefreedom]

32. i think this time

they are pretty jazzed.

if you check out pics of his events, lots of young people.

[bobclark86]

88. Compared to Obama? No n/t

[restorefreedom]

137. have you seen pics of his events?

majority are millenials
and who is spreading the word on social media? It's the young ones.

[SwampG8r]

180. As the hillarians are quick to point out

Obama isn't running this time

[Bubzer]

260. Those are interesting claims. Where are your links to back your position?

[Erich Bloodaxe BSN]

51. One poster is claiming that they were polled, but were under a very small error range.

Either way, it's pretty skewed. Whether the numbers polled of those categories were '0' or just 'really close to '0', it's not much difference.

[Gormy Cuss]

277. No- skewed would be reporting the crosstabs for small cell sizes.

What's going on here is that the sample was set up to be nationally representative for the whole set of answers but not for every gradation of subcategory. Thus the distribution of responses is good enough with some weighting to report overall opinion but not good enough to slice it finely. Had they polled 1000 Dems using the same other demographic quotas the pollsters would have been able to report more data in crosstabs.

[hack89]

63. Wrong - you need to read the methodology section again

all those groups were polled. It specifically says so.

[Bucky]

190. I think most of those demogs should skew toward Clinton

- nobody under 50 (older voters favor Clinton)
- nobody without college-degree (more educated voters favor Clinton (sorry! but college educated also skews Republican))
- nobody from rural areas (probably not much of a factor among registered Democrats)
- no non-whites (that should skew to Sanders)
- no Independents (this only slightly to Sanders, as it includes as many moderates as lefties)

[Bucky]

215. Going to college does NOT mean smarter. It means more income, tho.

But oops, yes, it looks like my information is like 20 years old. Ha ha. Historically, a family being able to afford college had a strong correlation to being Republican. For instance, contrary to pop cultural stereotypes of Yippees and HardHarts, support for the Vietnam War cut strongly along formal education lines, with about 2/3s of college grads (a much smaller portion of the population then) tending to support the war in 1964-69 surveys.

As of the early 90s, registered Republicans still significantly outnumbered registered Democrats among those with Bachelor degrees, although those with graduate degrees were just as strongly pro-Democratic (both splits were at about 60-40). But in the Clinton years, the trend largely reversed direction, with college education starting to trend Democratic. Remember that, while the knee jerk response is "Oh, smart people are just naturally gonna vote Democratic" remember the Ben Carson rule: even people with post graduate degrees can believe in fucked up shit. Or, more helpfully, think of all those evangelicals with their business degrees climbing the corporate ladders and donating money to Pat Robertson, Mitt Romney, and Jeb Bush.

If you go back to 1992, when California was just starting to become a Democratic-leaning state, the Democratic Senate candidate Boxer, got 50-44% of all voters, but only an advantage of 51% to 43% among college graduates. http://www.cnn.com/ELECTION/1998/states/CA/polls/CA92SH.html

In his elections, Dubya actually managed to reverse that education trend briefly. But in 2008 Obama flipped the trend back ( http://www.wiredtoshare.com/estelles/the_strained_correlation_between_school_and_political_affiliation ).

But the correllation is not as strong as we'd expect: http://elections.nytimes.com/2008/results/president/exit-polls.html
In 2008 Obama got 51% of those with "some college education" and 53% of those with "college degree or more" formal education. Compare that to his overall cut of 53% vs 46% for McCain. In other words, it's a demographic preference that barely registers above the overall population trend.

Another useful article.
http://www.people-press.org/2015/04/07/a-deep-dive-into-party-affiliation/

The bias comes in correlating education to wisdom (or even to clarity of self-interest). College taught me to think critically, but I don't have any illusions that a lot of folks I knew back in my salad days matriculated without questioning how the economy or global diplomacy really works. Them right there are your Republican voters.

[DemocratSinceBirth]

10. The sub sample is too small to be broken out.

eom

[Ichingcarpenter]

12. Their conclusions are statistically irrelevant

since they polled no one under 50

[DemocratSinceBirth]

28. They polled people under fifty. The sample was just too small to be broken out.

More importantly the findings confirm the findings of other pollsters in the field at the same time:

http://www.realclearpolitics.com/epolls/2016/president/us/2016_democratic_presidential_nomination-3824.html#polls


In fact it's a tad bit bearing for Hillary Clinton.

[roguevalley]

79. that could mean 1 person over 50. crap poll

[pnwmom]

225. The numbers were significant enough to lower Hillary's numbers from the

50 and up scores every time. Something was lowering Hillary's overall score significantly, and that was the scores she must have had with the under 50's.

Take a look at the numbers again and you'll see.

[demwing]

82. Nonsense polling

CNN is basically saying:

"We didn't poll enough people under 50, therefore the MOE was extremely high. So, instead of polling more under 50 voters, we just ignored them."

Shameless...

[TheKentuckian]

267. What is the difference for practical application? The distinction seems to make no difference.

[pnwmom]

224. Finally! Someone who actually read the report. Thank you

for restoring my faith in DU.

From page 5:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

[stevenleser]

234. Yep, this is not an uncommon thing in polling. nt

[cascadiance]

11. Makes you wonder what percentage of phone no's polled were land lines...

... versus mobile numbers too.

[Ichingcarpenter]

15. Page one gives you that

Interviews with 1,006 adult Americans conducted by telephone by ORC
International on September 17 - 19, 2015. The margin of sampling
error for results based on the total sample is plus or minus 3 percentage
points.
The sample also includes 924 interviews among registered voters (plus
or minus 3 percentage points).
This sample includes 606 interviews among landline respondents and
400 interviews among cell phone respondents.

[virtualobserver]

55. 60% landlines- I barely know anyone who has a land line

and the ones that I do know are all over 65

[WinkyDink]

57. Guilty as charged. Is it a moral offense now to have a landline? Do I live an inferior lifestyle?

I have a tablet, so can I at least watch the cool kids play "Candy crush"?

[liberal_at_heart]

60. not a moral offense to have a landline. Just shows how narrow the sampling is. That's all.

[virtualobserver]

85. The point is that a larger percentage of the younger population do not have land lines

There is a secondary category of people who have land lines and do not answer them.

[mythology]

108. As somebody else has pointed out, 60% is fairly accurate

As for the potential of that causing the poll to skew away from younger voters, younger voters (I am one myself) don't regularly vote. As such undersampling them as compared to their percentage of the overall population, as long as it's in line with their historical voting percentage is fine.

[virtualobserver]

112. only fine as long as the sampling represents the underlying reality

when the assumptions that must go into every poll suddenly do not reflect reality, then you get interesting surprises.

[liberal_at_heart]

118. Yes, there are lots of surprised people when actual voting results don't match polling.

Not so surprising if you realize how skewed polling is.

[liberal_at_heart]

116. No it is not okay to leave groups of people out of polls. I am 39 and don't have a landline.

I am not one of the youngsters that never votes. It is not okay to leave my age group out of it. And by the way college kids are waking up to the fact that they are gathering more college debt than they will likely be able to pay for and only part time and low wage jobs waiting for them when they do graduate. They are becoming extremely frustrated and are starting to pay more attention. I think the more debt they accrue the more young voters we will start seeing actually voting. Things are changing fast in this country. You can't necessarily count on things happening just as they always have been.

[jeff47]

78. 60% actually fits the overall statistics

At least, according to this article from 2014: http://www.chicagotribune.com/business/chi-landlines-survey-20140708-story.html

[virtualobserver]

98. it does, but younger people are unlikely to have land lines

and there is a third group that has "unanswered" land lines.

there has been a steady movement away from land lines

[jeff47]

102. True. (nt)

[Elmer S. E. Dump]

100. 60 / 40 in favor of land lines

[DCBob]

13. That makes zero sense

I think its just a typo.

[DemocratSinceBirth]

22. See Post 7

First of all it is a random sample so that means every member of the universe to be sampled which in this instance is Democratic primary voters has the same theoretical chance of being polled.

Also, just start willy nilly calling people on your phone. You aren't going to reach four hundred people and none of them will be under fifty years old.

[Ichingcarpenter]

25. The poll is worthless

I stand by that statement after looking at their data

[DCBob]

26. Yep.. also see my last post.

Sept 10 CNN poll had same.

[DemocratSinceBirth]

34. That poll is actually a little bearish for Ms. Clinton

That poll is actually a little bearish for Ms. Clinton. She is doing a bit better in the RCP Poll Of Polls:

http://www.realclearpolitics.com/epolls/2016/president/us/2016_democratic_presidential_nomination-3824.html#polls


and the Huff Po Poll of Polls:


http://elections.huffingtonpost.com/pollster/2016-national-democratic-primary


Both RCP and Huff Po are using some variation of the Law Of Large Numbers; by averaging polls or samples you get a bigger poll or sample and consequently more robust or meaningful results.

[DCBob]

35. Agreed... just noting that they also didn't include % of respondents under 50.

[DemocratSinceBirth]

43. They did. The sample was small to break out.

Similar to throwing Asians, African Americans, Hispanics, and Others into one big subgroup called "non-white".

[demwing]

77. In other words...

"We didn't poll enough people under 50, therefore the MOE was extremely high. So, instead of polling more under 50 voters, we just ignored them."

Shameless...

[DemocratSinceBirth]

83. Let's throw out that poll and use the other polls of recent vintage:

Let's throw out that poll and use the other polls of recent vintage:



http://www.realclearpolitics.com/epolls/2016/president/us/2016_democratic_presidential_nomination-3824.html#polls



http://elections.huffingtonpost.com/pollster/2016-national-democratic-primary



I actually like them better

[artislife]

167. Yup

I let unknown numbers go to voice mail, especially if the area code is strange. or 800 type.

[pnwmom]

226. You're right, and it's untrue. The OP thinks N/A means that no one was in

that category, and this is what it actually means.

"Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A"."

[randr]

16. Pollsters are paid for their work

and they are paid by the people who want the polls to be slanted toward them. The "science" of polls is to return the results you are looking for.
This is the reason that in the final analysis the polls are generally wrong.

[DemocratSinceBirth]

20. Here is a link to the Real Clear Politics "Poll of Polls"

http://www.realclearpolitics.com/epolls/2016/president/us/2016_democratic_presidential_nomination-3824.html#polls

The findings from the most recent CNN poll confirm the findings of other polls taking during roughly the same time period. The Law Of Large Numbers suggest as your sample grows your results become more robust, i.e. meaningful.

[restorefreedom]

21. its all bullshit

and cnn is pro corporate so we know what that means

bernie and his supporters are fighting a megaoligarchy with a lot to lose

DID WE REALLY THINK THEY WOULDN'T PULL THIS SHIT??

its kind of good news

if they have to rig the polls to get Hillary even competitive, things are way better for Bernie than we thought.

[LonePirate]

23. People under 50 were polled.

Compare the numbers for 50+ with the total/final numbers and you will see the impact of the under 50 crowd. I'm not saying the 49 and under numbers are large; but they certainly exist in this poll, even if they are not included in the age demo breakdowns near the end of the report.

On page 11, Clinton shows 57% total but 64% for 50+. Sanders has 28% total but 21% for 50+. If no one over 50 was polled, those pairs of numbers would be identical (ie Clinton's total and her 50+ number would be the same). The fact that the pairs of numbers do not match suggests (proves?) there were significant numbers of under 50 respondents.

[questionseverything]

65. depends on what you call significant

7% of 329 dems or indys leaning dem were under 50...so we are talking 28 peops?

since we do not elect presidents nationally the poll is meaningless anyway

[liberal_at_heart]

103. exactly. You can skew a poll to make it say anything you want.

[pnwmom]

227. Exactly! Thank you for getting it right.

And there is also this explanation in the report. The "N/A" notations did not mean there were no respondents in the category, as the OP and others seem to think.

From page 5:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

[DCBob]

24. The Sept 10th poll had same issue.

Page 30..
http://i2.cdn.turner.com/cnn/2015/images/09/10/demsclinton.pdf

Not sure why but seems like a non-issue to me.

[dsc]

27. someone is cooking the books

but it isn't CNN. It says on page 5 that crosstabs were only included if the MOE was under +/- 8.5. Page 6 has crosstabs for ages you say they didn't poll. The didn't poll enough of that category to keep the MOE under +/- 8.5 but they did poll some.

[JackInGreen]

30. Oh that's rich

I cone in to see the Hillionares flag waving, figured it'd be cooked books.

[roguevalley]

84. it is. crap as usual. they are trying to reassert

Inevitability

[PatrickforO]

33. Somebody's on the gravy train doing polls like this...

Waste of money for whomever paid for it.

[in_cog_ni_to]

36. So that's how they plan to keep Hill's numbers up.

We should have known have known they'd even cheat in polling.

There's now more Millennials than Baby Boomers and Millennials do NOT support Hillary.

[Enthusiast]

263. +100%!

[Fawke Em]

37. Perception, perception, perception.

CNN looks to be "all in" for the next anti-Sanders meme: "I like him, but he can't win."

[INdemo]

38. Sssssssssshh

Don't say too much about that you are spoiling all those wet dreams of the corporate media....

[Betty Karlson]

39. Most trusted name in news, everybody!

Even the onion is more reliable.

[berni_mccoy]

40. AND Sample Size was 392 which yields +/- 5% error

[Pope Sweet Jesus]

41. Is that the poll where every Clintonite DU'er

claims that Clinton is rising?

Bad poll! No soup for you!

[Erich Bloodaxe BSN]

52. I think it was a poll OF every Clintonite DUer ;) nt

[Hiraeth]

56. I am wondering the same thing. I think it is.

[Travis_0004]

42. Page 6 clearly shows that they did poll from all age group

Interviews were conducted among these subgroups, but results for groups with a sampling error
larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

[DemocratSinceBirth]

47. If we could resurrect Elisabeth Kübler-Ross she would say some folks are still in the Denial Stage.

Here is a link to the RCP Poll of Polls and the Huff Po Poll of Polls:

http://www.realclearpolitics.com/epolls/2016/president/us/2016_democratic_presidential_nomination-3824.html#polls



http://elections.huffingtonpost.com/pollster/2016-national-democratic-primary

[kenn3d]

75. Ummm... please look again.

The RCP composite does now include the new CCN poll, but the reported spread has actually decreased by 1% to Clinton +19.5. Not sure why yet... ? (This indicates that the lead she has over Sanders is still decreasing, but that's not what I would expect from the raw results).

The HuffPollster composite has not yet included the new CNN results (as of this writing). Stay tuned.

There seems to be some doubt about the validity of the new CNN poll samples and/or methods, but I'm inclined to believe that they are still among the most reliable pollsters, and if anything this poll might be something of an outlier due to normal random sampling not producing sufficient data samples for an acceptable MOE in several subcategories. It happens.

In any case, this one national poll so far out, still tells us very little beyond how many people know who Sen Sanders is.... So nobody in either camp should declare victory or defeat based on it.

Peace

[OilemFirchen]

89. Biden's numbers are up. (n/t)

[DCBob]

105. Has anyone declared "victory or defeat" based on this poll or any other single poll??

That would be ridiculous.

[liberal_at_heart]

53. oh, so every other group just happened to have a sampling error larger than

+/-8.5 percent? Pretty fishy.

[stevenleser]

235. Nope, not fishy. Common issue with polls when a small sample size is broken down further. nt

[Skwmom]

44. So under 50 were polled? n/t

[pnwmom]

228. Yes , they were, which is why all the overall results for Hillary were LOWER than the

results for ages 50 and over. People under 50 were pulling her scores down.

The OP mistakenly thought there were no respondents under 50 because those groups were labeled, "N/A." But this is what N/A means:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

[NorthCarolina]

48. It's a poll with an agenda. eom

[wilsonbooks]

58. I took a poll

And all my friends and family will be voting for Bernie. I did hear that there was an elderly couple in the next town over that are supporting Clinton, they must have called them.

[treestar]

59. Lie; they were polled, but their subgroup not broken out.

[DemocratSinceBirth]

64. To paraphrase Upton Sinclair...

To paraphrase Upton Sinclair never try to get somebody to believe something their whole world view hinges on not believing.

[BeanMusical]

93. Lol, the irony.

[DemocratSinceBirth]

104. I am probably the most literal person on this board and the person's whose real life persona...

I am probably the most literal person on this board and the person's whose real life persona is most like his internet one.

KISSES

[BeanMusical]

106. Love you too!

[DemocratSinceBirth]

110. Have a terrific week

eom

[BeanMusical]

113. Buh-bye, literal man!

[merrily]

136. Sinclair said "salary," not world view.

[PosterChild]

230. S/he said it was a paraphrasing nt

[merrily]

266. That was not a paraphrase of what Sinclair said, though.

A paraphrase does not change the meaning of the original statement, only rewords, usually to make the original statement more clear.

[PosterChild]

273. Thanks for the tip eom

[hack89]

62. Wrong - go read the methodology description again

you will see your error there.

[think]

72. "Some subgroups represent too small a share of the national population to produce crosstabs

with an acceptable sampling error."

Methodology

A total of 1,006 adults were interviewed by telephone nationwide by live interviewers calling both landline and cell phones.
Among the entire sample, 26% described themselves as Democrats, 23% described themselves as Republicans, and 51% described
themselves as independents or members of another party.
All respondents were asked questions concerning basic demographics, and the entire sample was weighted to reflect national
Census figures for gender, race, age, education, region of country, and telephone usage.
Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of
+/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with
an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error
larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

http://i2.cdn.turner.com/cnn/2015/images/09/20/dempoll.pdf

[Fawke Em]

96. So, basically, they still didn't poll very many people under 50.

That really doesn't change things.

Maybe they polled 1. Maybe they polled 10. Who knows?

The point is that they've over polled people over 50.

[think]

120. Yep. CNN chose not to show a breakdown by age group which is telling.

It is obvious that they didn't poll a significant number of people under 50 for whatever reason....

[DanTex]

71. Of course they polled people under fifty, they just don't have the breakdown.

Notice that the full population results are different than the over 50 results on page 7. If they only polled over 50, they'd be the same.

[DCBob]

97. Correct obviously.

Amazing how clueless people become when they don't like a particular result.

[Mass]

73. They were polled. The subsamples were too small to be significant.

CNN has a very small sample to start with. Numbers on the subsamples are mostly not significant, and they made the choice not to post those where the MOE was higher than 8.5 %.

The CNN poll has a MOE that is 5 %, which makes the results hardly reliable (same for their Republican poll)

[riversedge]

94. " makes the results hardly reliable (same for their Republican poll)"

[fredamae]

87. Reminder

Polls are only as reliable/credible as the person/group paying for them is honest. Polls are "gerrymandered" in both script and geography in order to elicit a particular desired response.

And this goes across the board...including the ones we Hate and the ones we Love.
"Trust but Verify" all
Just my opinion.

[DemocratSinceBirth]

92. I once was on a sports board and a poster wrote that Shaquille O'Neal ...

I once was on a sports board and a poster wrote that Shaquille O'Neal had a better three point shooting average than LeBron James because he misplaced a decimal point.

It was something like this:

Lebron- .325
Shaq- .0325


His mistake was pointed out to him and he eventually admitted he was mistaken.


There is a moral to that story.

[Android3.14]

101. That's how you keep the veil of legitimacy in polling

You can pay pollsters for any results you want. All you have to do is filter the population down to a pool of individuals from which you can then draw "random" responses.

[demwing]

109. Perspective: # of registered voters in the US

http://www.census.gov/hhes/www/socdemo/voting/publications/p20/2014/tables.html

According to the official record, there were just over 142 million registered voter in the US as of 2014. Of those, about 54.7 million are younger than 45 (the data groups break at the 45 year old mark, not 50).

That's upwards of 38.5% of registered voters who were ignored in this poll.

And yet CNN's excuse was "Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error"?

38% is too small?

More shameful behavior from those who know no shame...

[unblock]

111. well, of course! they said "adults"!

[Scuba]

114. Many (most?) polls are designed to influence opinion, not measure it.

[liberal_at_heart]

121. +1000000

[still_one]

127. This is a poll of registered Democratic voters. It is pointing out that in this sampling there were

not many registered Sander's supporters. To ignore it would be very foolish, and it would be prudent for Bernie supporters to get out a registration effort

[workinclasszero]

115. If Bernie is depending on the youth vote...he's screwed

Issued April 2014

Young-Adult Voting: An Analysis of
Presidential Elections, 1964–2012
By Thom File


Voting rates, which represent the
number of voters relative to a given
population or subpopulation, have
varied across recent election cycles
with the general outcome being
that voting rates increase with age.

In every presidential election since
1964, young voters between the
ages of 18 through 24 have consistently
voted at lower rates than all
other age groups, although young adult
voting rates have fluctuated
from one election to another.

Overall, America’s
youngest voters have moved
towards less engagement over
time, as 18- through 24-year-olds’
voting rates dropped from 50.9
percent in 1964 to 38.0 percent in
2012.

The decline in voting rates
discussed in this first section is
partially due to the increase in the
noncitizen population, which by
definition does not vote.


https://www.census.gov/prod/2014pubs/p20-573.pdf

[still_one]

122. The poll was of REGISTERED Democratic voters, and what it implies is that Sander's supporters have

not registered to vote in significant numbers.

A lot of people on this thread are laughing about it, but at their own peril

[noiretextatique]

243. it could also mean no sanders supporters were contacted

for the poll.

[still_one]

117. The key word is REGISTERED Democrats. Those who are not registered were not included in that series

that involved who they would prefer as the nominee.

In the question regarding Approval For President Obama, registration was not a factor, and they were included:

Where:

Do you approve or disapprove of the way Barack Obama is handling his job as president?

18-34 35-49 followed by other age groups
55% 36%
36% 54%

For those Bernie supporters who choose to laugh at this poll, I suggest you it would be in your interest NOT to ignore it, and get as many Bernie supporters registered as possible, because what the poll indicates is those Bernie supporters haven't registered to vote, and you cannot vote unless you are registered

[George II]

123. The document says "N/A", if there were none it would say 0%

The document probably had a bug when converted to the PDF. There are lots of illogical "N/A"s in the tables. Look at the next page where they talk about the location of those polled - "N/A" for Northeast, Midwest, South, West. So where did they poll, London? I doubt it.

Look for a modified release with complete data.

Easy to jump on something and not try to understand that there's something wrong with the document.

[BeanMusical]

125. I wonder if they polled DWS and all her DNC friends too?

[merrily]

126. LOL!

[DCBob]

130. Here's the key comment from the methodology..

"All respondents were asked questions concerning basic demographics, and the entire sample was weighted to
reflect national Census figures for gender, race, age, education, region of country, and telephone usage."

[still_one]

135. The question regarding the Democratic candidates involved only registered Democratic voters

[liberal_at_heart]

138. There is no way to rationalize leaving out millions of people in a poll.

It will only lead to the polls being wrong. You can't possibly predict how an election will turn out when you leave that many people out of the polling. Someone said that polls are designed to influence voters, not monitor them. I believe that. It is a marketing tool that campaigns use to discourage voters and influence them to switch their vote. There is no way in hell I am switching my vote, not even at the caucus. I remember my very first caucus. I was stunned at how much bullying goes on to get people to switch their vote. The way we hold elections in this country is so screwed up. It's no wonder so many people don't vote.

[DCBob]

140. They were not left out.

They were just not displayed in the crosstabs section due to large error rate. The overall survey result includes them. This only impacts the crosstabs breakdown. Its a survey science thing... if the error rate is very high don't even show it because it could be misleading. This is a good poll.. imo.

[liberal_at_heart]

143. Why was the error rate so high? How many people over 50 did they poll versus people under 50?

Never mind. I'm done arguing. You can justify it any way you want. Many of us know better.

[DCBob]

150. Do you actually believe any pollster would conduct a poll leaving out half the population??

Ridiculous.

[demwing]

156. "Ridiculous"

Yes, the poll is ridiculous

[demwing]

152. THIS is the correct question

Why was the MOE so high?

According to the poll, it's because the under 50 group was not sufficiently polled.

The pollster had two choices - Do more polling to rectify the error, or just say "fuck it" and toss the results.

Apparently, they took the "fuck it" route...

[NuclearDem]

142. This thread is a good argument for mandatory college-level statistics courses.

N/A doesn't mean nobody was polled.

[demwing]

145. It means the MOE for the under 50 group was too high to include

which means that not enough people under 50 were polled, despite the fact that they make up about 40% of registered voters

[liberal_at_heart]

151. Exactly. Thank you.

[NuclearDem]

175. Which is a common problem with random sampling.

Under 50 might make up 40% of the population, but the results of random sampling might not reflect that.

[DemocratSinceBirth]

192. "Under 50 might make up 40% of the population."

Under 50 might make up 40% of the population.

Actually, we are looking for those 18-49 as those under 18 can not vote .

[druidity33]

222. Actually that's around 40% of REGISTERED voters. nt.

[liberal_at_heart]

149. How many people in each age group were polled?

Are you telling me you can't skew poll results by making the margin of error what you want it to be by only sampling a few people in one group and a lot of people in the other?

[sabrina 1]

247. What it's a good argument for is that political operatives trying to skew results are going to be

caught because we don't live in the dark ages anymore.

All these attempts to 'control' the 'message' are failing. Which is a good argument for just being HONEST.

[Trajan]

146. Fuck the media

We've gotten this far through sheer force of will ... Through a message that resounds and a messenger that garners respect for his courage and integrity ...

It's going to be a long slog ... This is just beginning, so push forward with confidence ...

Bernie WILL win ...

[questionseverything]

158. in a way the poll is good news for bernie

shows the ptb are past the ignore,ridicule stage and are at the fight stage

[stillwaiting]

216. Yes. This. nt

[Beartracks]

147. I sure hope the younger folks are getting registered.

Not only will non-registered people possibly be ignored for certain polls, but they will most definitely be ignored on Election Day!

===============

[drm604]

154. On page 6 of that PDF it shows a breakdown by age, starting with the 18-34 range.

The lowest age group is 18-34.

It shows these breakdowns: 18-34, 35-49, 50-64, 65+. It also breaks it down into two groups; under 50 and over 50.

Am I misunderstanding what you're saying?

[DemocraticWing]

157. I'm sorry but you got it wrong

Polls frequently don't include crosstabs of subgroups, especially the smaller ones like 18-34, nonwhites, etc. who are lower parts of the sample because turnout among those groups is relatively very low. When you have such a small subsample, the MOE is so high that the number is basically worthless for that specific group.

The polls goes so far to say this on Page 5.

I think the poll seems about right anyway, Clinton's been in the mid 40s for a while and Sanders has been in the Mid 20s for a while. Both numbers dipped slightly as Biden's support has ticked up.

[questionseverything]

163. technically the op is wrong but a sample of 21 or 22 voters for the under 50 set is worthless

329 dems or dem leaning indys were polled

there is a difference of 7% from hillarys total support to her over 50 support,same with bernies

so 7% of the 329 were under 50

which is 21 peops

national polls are meaningless anyway since we elect presidents state by state

but it gave cnn the headline the ptb needed

[OilemFirchen]

177. Wow.

Seriously.

Wow.

[muriel_volestrangler]

159. They polled between 90 and 114 Democrats/Democratic leaners under 50

When they do give a sampling error, that allows you to put the sample size in a range, given from these figures: http://www.lifestylesurvey.org.uk/05_methods/sampsize.html (and the general formula is SS = 10000/(SE*SE)

They quote the confidence interval to the nearest 0.5%. They tell us they polled 392 Democrats/Democrat leaners; with the 50+ group at +/- 6.0, that's up to 302 in that age group, so 90 or more under 50. 65+ is +/- 8.0, so that's at least 147 polled; 50-64 is +/- 8.5, so that's at least 131. So the most of 'under 50' they polled was (392-147-131)=114.

They say they weight the results according to census figures. We have to take their word that they've done that correctly. What this does show, however, is that they didn't get very many answers from people under 50, so it may not be that accurate a poll of Democratic voters (according to the 2012 exit poll, about 60% of Obama's vote was under 50).

There is a difference between primary voters and those who vote for a party in a general election, of course; it's possible that those bothering to vote in primaries are older than in a general election. If anyone knows rough figures for that, it might help.

[liberal_at_heart]

161. Thank you for those numbers.

[questionseverything]

173. disagree

BASED ON 261 REGISTERED VOTERS WHO DESCRIBE THEMSELVES AS DEMOCRATS AND
131 REGISTERED VOTERS WHO DESCRIBED THEMSELVES AS INDEPENDENTS WHO LEAN
DEMOCRATIC, FOR A TOTAL OF 392 REGISTERED DEMOCRATS-- SAMPLING ERROR: +/- 5%
PTS.

if the moe is 8% then the under 50 is less than 8% or it would not be described as n/a

in my earlier posts i had transposed the 392 to 329 so my math was a little off but still .08 of 392 is 32 peops at .07 it would be 28 peops

the entire poll leans conservative since over a 1000 voters were polled and only 392 identified as dem/leaning

[muriel_volestrangler]

203. No, that's not how sample errors work

"if the moe is 8% then the under 50 is less than 8% or it would not be described as n/a"

No. As the PDF says, they write "N/A" for any group with a moe above 8.5%. That means a sample of 130 or under - see http://www.lifestylesurvey.org.uk/05_methods/sampsize.html .

[questionseverything]

211. look at page 11 in the poll

we know the whole number is 392

hc receives 57% overall but receives 64% in the 50 plus group

bs receives 28% overall but receives 21% in the 50 plus group

both exactly a 7% exchange...there are only 2 groups to look at under 50 and over 50, so what else besides the under 50s could account for the difference?

heck if cnn wanted us to know how many under 50 were polled they could of just told us...lol

they did not follow your pdf too closely because 392 should be 5.5 moe not 5 (thats 400)

anyways national polls are meaningless but here is a whole number break down

hc@57%=223
om@2%=8
sanders@28%=109
nobody/someone else/no choice @13%=50

[muriel_volestrangler]

214. What you say in #173 doesn't make much sense

so I'm having to try to interpret what you wrote, but:

"BASED ON 261 REGISTERED VOTERS WHO DESCRIBE THEMSELVES AS DEMOCRATS AND
131 REGISTERED VOTERS WHO DESCRIBED THEMSELVES AS INDEPENDENTS WHO LEAN
DEMOCRATIC, FOR A TOTAL OF 392 REGISTERED DEMOCRATS-- SAMPLING ERROR: +/- 5%
PTS.

if the moe is 8% then the under 50 is less than 8% or it would not be described as n/a "

I don't know where you've got your '8%' from. No, the N/A is used in the following situation - this is a direct quote from the poll:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of
+/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with
an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error
larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

They look at the sampling error that applies to each subgroup, based on its unweighted size. For 392, it would be the square root of (10000/392) = 5.05; they rounded that to the nearest half percentage point, ie 5.0%.

" .08 of 392 is 32 peops at .07 it would be 28 peops "

I can't tell what you were trying to say here. Neither 32 nor 28 turn up in your post after this, nor in the poll.

As for this latest post:

Yes, the under 50s are what makes the difference between the overall result and the over 50s result. Of course. But that doesn't tell us anything about the size of the under 50 sample. As someone has already pointed out, I think, " the entire sample was weighted to reflect national Census figures for gender, race, age, education, region of country, and telephone usage." Notice the quote above talks about the unweighted subgroups' size.

We can't tell, from this difference alone, whether there's a large weighting for the under 50s for which the figure wasn't too far off from the over 50 one (eg if it had equal weighting, you'd have HC at 50% for that question, and BS 35%, so that those figures are an equal distance from the overall candidate figures as the over 50s), or if it's a smaller weighting, but the differences were larger (eg if the under 50s were only one third of the total weighting, then their figure would be twice as far from the average - so that would be HC=43%, BS=42%).

But even if we knew the weighting, that wouldn't give us the unweighted sample size. The point is that the pollster has to make up for getting fewer answers than they'd like from younger people by making that weight up to the Census level.

"heck if cnn wanted us to know how many under 50 were polled they could of just told us...lol "

Yes, they could have, but they didn't. It's annoying. I did the calculation to try and find out what the number polled was.

[questionseverything]

217. no time now but can we agree on the whole numbers the poll produced?

hc@57%=223
om@2%=8
sanders@28%=109
nobody/someone else/no choice @13%=50

it only adds up to 390 but it is pretty close

[muriel_volestrangler]

219. If you want - those numbers have nothing to do with how many under 50s were polled

but if what you wanted was the number of people replying with each answer in that question, then I'd say that's right (after their weighting).

[questionseverything]

253. 25% of respondents were under 50

8moe 156
8.5moe 138
=294 over 50

392-294=98

98 =25% of 392 total voters

of those under 50 voters, 47 vote for bernie ,37 hc,14 others

i was mistaken earlier thinking it showed even lower but a 75/25 split is not a good sample

thanks for the lessons!

[stevenleser]

250. See her #159 again and my #239 above where I show the numbers of

Folks under 50 they would have polled if they used the census results.

[Sunlei]

160. looks like if Joe Biden became VP for Senator Sanders, that 'team' could win the primary

[Warren DeMontague]

166. BREAKING: "9 out of 10 Hillary Clinton supporters, support Hillary Clinton!"

[Warren DeMontague]

168. ...They also polled people with "aol.com" email addresses

Dont tell me that's not a representative sample, sonny!

[senz]

169. Grossly lopsided sample, worthless poll.

To give it any usefulness at all, they should have put the selected demographic in the heading. If you observe the "N/A" counts, you see that no one under 50 AND no non-college graduates were included.

This poll does NOT even come close to representing all Democrats.

[liberal_at_heart]

170. exactly. If they want to poll a certain demographic then say so. Don't make it sound like

you are sampling everybody when you are clearly sampling a certain demographic.

[DCBob]

172. How about this poll?

This is previous CNN poll showing Bernie closing within 10 points. It also did not show respondents under 50 in the crosstabs. Worthless also??

[liberal_at_heart]

176. if they only polled a couple dozen in one group and a couple hundred in another group then yes

it is worthless.

[DCBob]

178. Ok, so its not a conspiracy... just CNN is a crap polling outfit.

Is that correct?

[liberal_at_heart]

182. You were making a point about Bernie supporters complaining about that poll and not another.

Does that mean Hillary supporters won't complain if she wins the nomination and the Republicans run skewed polls against her? Polls can be and often are skewed based on who is asked and how the questions are asked. I think we all know that.

[DCBob]

183. Do you think CNN purposely skewed this poll?

[liberal_at_heart]

185. It is either a conspiracy or the entire polling system is screwed which is probably true which means

we can't count on any polls. If that is true why do we have polls? What is the purpose of having polls if the sampling size is so off? How can you have an accurate poll when you poll 20 people in one group and 200 in another? And what happens if those 20 people represent a group of people that make up nearly half of the population? How is that an accurate representation of the voters?

[DCBob]

187. I have to admit I do not know why they were not able to poll enough under 50 respondents..

to get a acceptable error level at each age group but I do know this can happen in surveys and they simply deal with it the way CNN did it. Weight the data but don't show the numbers in crosstabs. Its still considered legit.

[progressoid]

193. Yes, worthless.

[DCBob]

194. So throw out all CNN polls?? Seems this is their standard methodology.

[progressoid]

197. If that is their standard methodology, yes

18-44 year olds represented 45% of voters in 2014. http://www.census.gov/content/dam/Census/library/publications/2015/demo/p20-577.pdf

Would we want a poll that ignores all boomer voters?

[DCBob]

200. You still don't get it.

http://www.democraticunderground.com/?com=view_post&forum=1014&pid=1212426

[progressoid]

204. I get it. Large error rate. Small sample size.

A voting block that consists of nearly have of voters and they couldn't get a sample to produce an error rate better than -/+8.5%?

Maybe try harder next time.

[DCBob]

205. I dont think you understand its weighted to deal with that..

So its legit.

[progressoid]

208. Weighted for a small sample size.

The lower the sample size, the higher the margin of error and lower the confidence level.

I'm not denying they were included in the survey. Weighting a demographic that can't meet a MOE of 8.5% isn't very reassuring.

[DCBob]

209. That I will agree with you on.

But it does seem that is a somewhat normal procedure... at least for CNN.

[pnwmom]

229. Older people are more likely to be registered to vote. Young people can change the situation

by registering and voting in higher numbers.

http://www.huffingtonpost.com/2012/09/29/young-voters-fewer-are-registered_n_1924181.html

Young voters turned out in droves to help propel Barack Obama to the presidency in 2008. While Obama still holds a huge chunk of support from young people, fewer are registered and they're less engaged than older demographics, according to new polling from the Pew Research Center.

Only 50 percent of people ages 18-29 are registered to vote -- a full 11 percent lower than in 2008 and the lowest number in the past 16 years of Pew's polling. The number of young people registered to vote hasn't been this low since 1999. The Pew Research Center noted young adults are at their "lowest registration rate of the last five presidential elections."

The overall voter registration number currently sits at 72 percent. All other age demographics have at least 70 percent registered to vote, although each age group is down slightly from 2008 levels.


SNIP

[senz]

195. Yes. CNN should be held to account for their shoddy methodology and presentation.

In Bernie's case, DCBob, the truth would almost certainly have been even better for him, as a huge part of his following is with younger Democrats (as well as Independents).

[DCBob]

196. CNN is one of the highest ranked polling outfits by 538 with an A- ranking.

http://fivethirtyeight.com/interactives/pollster-ratings/

If you throw them out you have to throw a bunch more since I think their methodology is standard survey science.

[senz]

199. Well they screwed up here. As I said, it's either methodology or presentation or both.

They could have saved face by putting the sample demographic (or irregularities encountered) in the headline and in the report. As it is, they misled their audience.

BUT -- your assumption that I or anyone else is suggesting that CNN be entirely "thrown out" as a polling organization is simply jumping to conclusions.

They need to correct their error, apologize, and in the future pay closer attention to what they're doing and how they report their findings.

[DCBob]

201. Not a screw up.. seems just the way they do polling..

http://www.democraticunderground.com/?com=view_post&forum=1014&pid=1212426

[stevenleser]

252. Not correct. See #s 159 and 239 above. Nt

[David__77]

179. I don't think that's true.

According to the survey details located at http://i2.cdn.turner.com/cnn/2015/images/09/20/dempoll.pdf:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/-8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error.

What this does tell me is that there were more people surveyed in the 50-64 age group than in the 18-49 age group, and there were more people surveyed in the 65+ age group than in the 18-49 age group.

[DCBob]

181. I think you have that correct..

Then they adjusted (weighted) the data to reflect the real world.. Here's the statement from the methodology..

"All respondents were asked questions concerning basic demographics, and the entire sample was weighted to
reflect national Census figures for gender, race, age, education, region of country, and telephone usage."

This is normal survey science.

[David__77]

184. Reweighting the groups makes sense to me.

I do find it interesting that there were more 50-64 aged respondents than 18-49 aged respondents. Reweighting conceptually addresses the skew compared with the population. At the same time, the low statistical precision of the 18-49 group means that the actual preferences are relatively likely to be quite different from the sample preferences for that age group. Certainly that happens with small demographic groups in a survey population.

[liberal_at_heart]

189. Thank you such a thoughtful and informative post.

[DCBob]

191. Yes, that is not clear to me either.

I don't know why they were not able to poll enough from those age groups but as far as I know, the way they handled it, is legit.

[Bucky]

188. "sample includes 606 interviews among landline respondents and 400 interviews among cell phone"

Who answers their cell phone from an out of state caller they don't recognize? Just us dinosaurs. Yall whippersnappers are all too busy texting while you drive your mopeds over to your part time jobs.

Yes, this survey is tragically skewed.

[valerief]

202. Poll dancing again. Page 7 says it all. nt

[Puzzledtraveller]

210. They know how to cook em!

[GeorgeGist]

212. Under 50yr olds don't count according to this disclaimer.

page 5

Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

IOW useless poll.

[pnwmom]

232. Wrong. That is not what the disclaimer means. They obviously count

because the overall results for Hillary are always lower than her results for over 50. It is the results of people under 50 that are pulling her scores down.

But why are there fewer people under 50? Young people are much less likely to be registered.'

And this is a study of registered voters.

[SidDithers]

213. Shouldn't you be editing your OP...

Now that it's been shown that you're wrong?



Sid

[blackspade]

220. Nothing to see here, move along....

Damn kids!

[Segami]

221. K&R

Nothing here to be surprised at. They are panicking. It won't be long before the rest of the nation will experience 'that moment' of clarity & truth.

[pnwmom]

223. Can't anyone read? That is not what these results say.

Last edited Mon Sep 21, 2015, 09:22 PM - Edit history (1)

1) The study DOES include some respondents under 50.

N/A does not mean there were no respondents in that group. According to the report, it means that the margin of error in the group was larger than 8.5%.

But even without knowing that, obviously there were respondents under the age of 50, because the overall results for Hillary were always lower than her scores among people over 50. What was pulling her overall score down? People under the age of 50.

2) This is a study of registered voters and the younger age groups are less likely to be registered.


From page 5:

Crosstabs on the following pages only include results for subgroups with enough unweighted cases to produce a sampling error of +/- 8.5 percentage points or less. Some subgroups represent too small a share of the national population to produce crosstabs with an acceptable sampling error. Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".


http://www.huffingtonpost.com/2012/09/29/young-voters-fewer-are-registered_n_1924181.html

Young voters turned out in droves to help propel Barack Obama to the presidency in 2008. While Obama still holds a huge chunk of support from young people, fewer are registered and they're less engaged than older demographics, according to new polling from the Pew Research Center.

Only 50 percent of people ages 18-29 are registered to vote -- a full 11 percent lower than in 2008 and the lowest number in the past 16 years of Pew's polling. The number of young people registered to vote hasn't been this low since 1999. The Pew Research Center noted young adults are at their "lowest registration rate of the last five presidential elections."

The overall voter registration number currently sits at 72 percent. All other age demographics have at least 70 percent registered to vote, although each age group is down slightly from 2008 levels.

[progree]

231. Page 8 - Hmm, no Republicans, nor anyone from the Northeast, Midwest, South, or West either

Page 5 - Interviews were conducted among these subgroups, but results for groups with a sampling error larger than +/-8.5 percentage points are not displayed and instead are denoted with "N/A".

Oh, that's what N/A means.

[stevenleser]

240. Yep, OP didn't read what they linked to and made wild conspiratorial assumptions instead. nt

[progree]

246. Well, sadly, that's standard DU. The really aggravating part is OP's refusal to fix the OP

when all this is pointed out to him/her. That's a bit exceptional. Or maybe I'm just being too persnickety.

I don't know if it is exceptional or not that so many swallowed it hook, line, and sinker. Too often, people are willing and eager to believe anything, no matter how ludicrous, if it fits preconceived notions.

[stevenleser]

286. Hillarious! Thanks for posting! nt

[Gormy Cuss]

278. That's the crosstab that should be a giant clue that "N/A" doesn't mean nobody interviewed.

[rladdi]

248. I suspect the GOP primary polls are much worst.

[Gman]

255. LOL! There's going to be a lot of under 50 people saying

You mean I can't vote for Sanders unless I register as a Democrat???

[PatrynXX]

257. face palm n/t

[C Moon]

261. I was watching CNN today (at the gym—and I say that because I NEVER watch it at home)...

and CNN was talking about (of course) Trump and the other GOP candidates.
But whenever they DID mention the Democratic candidates, they only showed pictures/video of Clinton.

Then, they announced a segment on the Democratic candidates—and even then they only showed pictures of Clinton and Binden before the break.

Finally, they mentioned Sanders and showed his photo...but it was only to say that the latest "CNN poll" put Hillary further ahead of Bernie.

CNN is obviously no longer news, it's advertising.
And I believe (after seeing the skewed polls in the last presidential election) the CNN polls are controlled and contrived.

To me, CNN is closing in on FOX.

[Enthusiast]

262. That seems fair.