X2- CNN poll released on March 19 did not include a large enough sample of persons under 50

The quality of a poll is only as good as the quality of the sample participants.  One needs to examine the “Sample and Methodology detail” of any poll to ensure that the sample is representative of the normal voting population. Attributes such as the gender, age, income and race of the participants need to be properly represented, otherwise the results will be “skewed”.  Skewed is a fancy way of saying “tainted” or “contaminated”. The numbers presented might be “accurate” for the particular population that was sampled, but to consider them representative of the entire voting population or having much predictive value is wrong and silly. It’s a Math thing.

Turns out that the poll conducted by SSRS and released by CNN on March 19 contains some curious results.

Notice anything?  Hint: look at the columns for 18-34 and 35-49

See all those N/A’s in the columns for the 18-34 and 35-49 age groups? Those are kind of weird, eh? Turns out there were not enough participants surveyed in those two age groups to make the answers statistically significant.

While interviews were conducted among a representative sample of the adult population of the United States, results for subgroups with fewer than n=125 unweighted cases are not displayed and instead are denoted with "N/A" because they are too small to be projectable to their true values in the population.

Note, however, that there were sufficient participants in the sample who were over the age of 50. A survey given to a population that has a lot of older voters, and a small amount of younger voters is clearly going to skew in a direction towards the preferences of older voters.  

SSRS claims that the subjects were “a representative sample of the adult population of the United States”. What does those words even mean? They provide no metrics about that attributes of their sample subjects, other than by political party.  Note that a recent poll by Quinnipiac University uses the following language instead:

The overall adult sample is weighted to recent Census data using a sample balancing procedure to match the demographic makeup of the population by region, gender, age, education and race.

Now that’s the kind of language that one wants to see in a scientifically valid poll: a sample population that actually matches the demographics of the voting population.

The CNN poll erased the opinions of younger voters from it’s results!  And yet CNN went on to make claims about the results, as if those results were actually meaningful.  Harry Enten published an article with the breathless title, Polls show Bernie Sanders popularity among all voters is plummeting!  Inside the piece, the claim is a bit more demure:

One of the arguments Bernie Sanders' fans made during the 2016 Democratic primary was that he was more electable than Hillary Clinton. His favorable ratings with the general electorate were far higher than Hillary Clinton's. Indeed, Sanders maintained fairly high favorable ratings with all voters as late as 2018.

Sanders' popularity among all voters seems to be declining considerably in the last few months, however.

Gotta love that word “seems”, lol. 

Enten cherry picks statistics from the survey in order to paint Sanders prospects as dim. He states that Sanders favorability “among all voters” has dropped to 46%, but he is comparing apples and oranges when he compares a number from a poll with few younger voters to one that matches the demographics of the voting population.  Also, he neglects to mention that in the very same survey, Sanders favorabiilty among Democrats and Independents who lean Democrat is a whopping 74%!  And that is with a sample skewed toward older voters! These are the voters who are going to choose who the Democratic nominee is going to be, of course.  The next most popular candidate from the survey in this same category is O’Rourke with 46%, followed by Booker at 44% (ratings on Biden, Harris, and Warren were not obtained).

While writing this essay, it occurs to me that our Daily Kos straw polls probably don’t capture the voice of many young people either. I believe that our community skews older, whiter, and richer, though perhaps I am wrong about that. 

Some wit of the past, I think it was Mark Twain, talked about “lies, damn lies, and statistics”.  Sadly, numbers can be used to mislead rather than to enlighten; when they are “presented” without any explanation of how the numbers might be skewed, innocent people tend to believe what they are shown.  I think that Twain would not be overly fond of this latest CNN poll.

Postlogue — A Visit to CENSUS.GOV

I wandered over to census.gov to look at the distribution of the entire US population, and found some interesting numbers (under Median Age and Age by Sex: “Annual Estimates of the Resident Population for Selected Age Groups by Sex: April 1, 2010to July 1, 2017”).

Persons in the 18-34 and 35-49 age brackets make up 23.4% and 19% of the total population, respectively.  However, if you look only at those who are of voting age, they make up 30.3% and 24.6% of all persons.  30.3 + 24.6 = 54.9, so these two categories make up over half of all potential voters.

2016- Total Population age population % 0 — 18 18 — 34 35 — 49 50 — 64 65+ total 73,658,812 22.8 75,625,795 23.4 61,519,367 19.0 63,329,631 19.6 49,272,330 15.2 323,405,935 100

2016- Voting Age Population age population % 18 — 34 35 — 49 50 — 64 65+ total 75,625,795 30.3 61,519,367 24.6 63,329,631 25.4 49,272,330 19.7 249,747,123 100

Again, SSRS claims that the subjects were “a representative sample of the adult population of the United States”.   According to the US Census, the largest category of voting age persons consists of those aged 18-34, and yet somehow SSRS did not have enough participants in their poll to make statistically significant conclusions about the preferences of these potential voters.  Amazing.

And then Harry Enten actually uses this poll to falsely claim Polls show Bernie Sanders popularity among all voters is plummeting!

The bottom line is that this CNN poll is misleading at best, and false propaganda at worst.  Voters need to be on the watch for fake polls that are trumpeted by the MSM to make claims that try to build a narrative that is not actually supported by data.  Shame on CNN for releasing a biased poll; Harry Enten should be embarrassed by not catching this gross error made by SSRS.

Finally, remember that even in a poll that is highly skewed to older voters, Sanders favorabiilty among Democrats and Independents who lean Democrat is a whopping 74%!  That is very good news for Sanders, and the exact opposite of the false narrative pushed out by CNN.