2016 Postmortem
Related: About this forumLet's do math. 45% of CNN's sample was under 50 years old
Let's break this down.
1006 respondents total. 26% Democratic. So 262 respondents were Democratic.
Sample for Democratic nomination was only registered Democrats. There were 924 registered respondents total (92%). Assume same proportion for Democrats and you get a new sample of 240 people.
Doing some basic math based on percentages given for each nominee, the lowest possible split for under 50 voters is 107 vs 133 for over 50. Math is not perfect but it is pretty close for information presented. And lo and behold a sample under 110 has a MoE over 8.5%.
Can you argue the entire sample is too small? Feel free. But let's end the BS that no one under 50 was considered when it makes up almost half the sample.
cali
(114,904 posts)still uses only land lines?
SonderWoman
(1,169 posts)"This sample includes 606 interviews among landline respondents and 400 interviews among cell phone respondents." See page 1.
Travis_0004
(5,417 posts)I've been called for polls before on my cellphone (I usually just tell them im not interested)
I would assume the vast majority use cell phones as well as landlines.
The CNN poll used 400 cell phones and 606 landlines.
nichomachus
(12,754 posts)I don't take cell phone calls from anyone I don't know. I send them to voicemail.
If it's important, they'll leave a message.
Also, I was talking to some 30-something guys the other day. They said they rarely answer cell phone calls, even if they know who it is. "If they want to get me, let them text," one of them said to me.
Cali_Democrat
(30,439 posts)You should know this.
Godhumor
(6,437 posts)They covered it in the demographics page for the poll.
PotatoChip
(3,186 posts)After all, a previous CNN poll excluded non-whites, and I doubt that there were not significant numbers of non-whites reached.
CNN may have asked the polling firm to specifically poll registered voters aged 50 and over. Polling firms are frequently asked to zero in on various demographics for (whatever) reason.
Godhumor
(6,437 posts)It is not that hard to figure out from the info presented what the approximate sample breakdown happens to be.
PotatoChip
(3,186 posts)you are not understanding me... So please correct me if I'm going down the wrong path...
The polling firm very well may have been asked to politely disengage from an otherwise willing participant -if- he or she did not meet a set criteria. In this case, being 50 or older.
This sort of thing (asking that a certain demographic be polled while excluding another) is not at all uncommon.
Godhumor
(6,437 posts)Because almost half their sample is under 50 years old. It is a weighted poll, so there may be a disengagement point for certain demographics, but it wasn't zero. It wasn't even close to zero. It was 45% under 50 to 55% over 50.
PotatoChip
(3,186 posts)Where did you get the info that 45% were under 50?
I'm not at all saying that I don't believe you; I just may have missed that. I'll go back and check page 7, but if that info is elsewhere, could you please let me know where it is?
Godhumor
(6,437 posts)Backfilling to get what numbers work based off percentage size for over 50 versus final percentage and then calculating results until I reached the minimum numbers needed for the MoE by subsample to work. I am confident I am within 2% of reality based on the assumptions I made in the OP.
I do this for a living.
The simple check is easy. If the poll was only 50 and over then the percentage for each candidate should match the 50 and over results. They don't. That means that the NAs contribute. In Clinton's case, she is 6% lower than her 50+ result.
PotatoChip
(3,186 posts)Godhumor
(6,437 posts)PotatoChip
(3,186 posts)So I see what you were saying now... My apologies.
Still, my guess is that younger people were less likely to agree to a (possibly) lengthy poll. It does not necessarily mean that they are not registered voters as some in a different thread suggested.
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"
restorefreedom
(12,655 posts)yup. like to reinforce a particular narrative, for example.
progressoid
(49,964 posts)Regardless of the demographic or pollster involved.
jeff47
(26,549 posts)If the poll more-or-less says "We haven't a clue what under-50 people think" by not including it in the crosstabs, why include those same people in the headline?
(Other than the need to publish any results at all, and a "change" result gets you more eyeballs)
If you don't trust your results in one part of the poll, it doesn't seem like a good idea to trust them in another part of the poll. Yes, you can mathematically make it "good enough" by diluting the accuracy of your "over-50" results.
dsc
(52,155 posts)LGBT is almost never cross tabbed for this reason. For my money they should have given the percentages, noted the MOE, and moved on.
jeff47
(26,549 posts)But "Under 50" is a rather enormous group of people. And they couldn't get enough under-50 respondents to make it work.
That doesn't exactly make the poll look like a useful predictor, especially with all the previous polls showing a large difference in support based on age. Even if you can get the MOE down far enough by piling everyone together.
dsc
(52,155 posts)the sample itself was OK. Piling everyone together is how sampling works.
Godhumor
(6,437 posts)They just didn't bother to parse out subsamples that individually have too large of an MoE. Part of these issue with this is that the full sample is small, though meaningful, which makes starting conclusions on subsamples extremely hard.
And CNN didn't include it in the headline. There is nothing saying "under 50s preferred x by y amount" because they do not feel they can make conclusions on the under 50 population...even if the subsample is in line with the full population parameters.
SidDithers
(44,228 posts)Thanks for adding some reality to the situation.
But that other thread was freakin' hilarious.
Sid