What Is The Science Behind Polls? — Transcript

Full transcript

1. 0:0096% of people like cheesy opening jokes according to a survey of DNews writers.
2. 0:07Hello voters, Trace here for DNews.
3. 0:12In this election year with so much stock being put into what the polls say, it’s important
4. 0:17they’re as accurate as possible.
5. 0:18But depending on the poll the results could be a reflection of how the nation feels as
6. 0:22a whole, or a bunch of fringe supporters skewing the outcomes.
7. 0:26What’s the science behind designing a poll that will accurately predict how the country
8. 0:31will vote?
9. 0:32Obviously, no poll can get ahold of everyone in the country, otherwise our phones would
10. 0:36be ringing constantly.
11. 0:37Instead pollsters have to interview a relatively small number of people and extrapolate their
12. 0:42data from there -- they call this a sample, or N. Scientific polling organizations will
13. 0:47aim to survey 1,000 to 1,500 people.
14. 0:50Statistically, the result will almost always have a 3% margin of sampling error then meaning
15. 0:56that these people should be representative of the population’s attitudes give or take
16. 1:003%.
17. 1:01They could survey more people, but after 1,000 people the margin of error doesn’t decrease
18. 1:05much.
19. 1:06To get the margin of error down to 1%, pollsters would have to survey 10,000 people.
20. 1:11At that point it’s not really worth the time or effort, not to mention money.
21. 1:15There are other kinds of errors that go unaccounted for, like errors in analysis -- where certain
22. 1:19assumptions the researchers make about how to handle their data turn out to be wrong
23. 1:23-- and the nonresponse error -- where people who support a candidate who’s behind are
24. 1:28less likely to respond, further skewing results.4,
25. 1:32Polling organizations like the Pew Research Center for example will find participants
26. 1:35by calling them, selecting the geographical area by area code and the next 5 digits of
27. 1:41a phone number.
28. 1:42Then they will randomly dialing the last two.
29. 1:45The phone numbers randomly selected from one area are supposed to be proportional to that
30. 1:49area’s share of phone numbers in the country.
31. 1:52The idea is to get totally random representatives from across the nation on the phone, though
32. 1:56with more people switching to cell phones it’s gotten a bit harder to pin down.
33. 2:02Getting them on the phone is the first step, but the researchers know they don’t have
34. 2:05an accurate sample just yet.
35. 2:07Older people are more likely to answer a landline, so the Pew Research center will ask for the
36. 2:12youngest male in the house half the time and the youngest female the other half.
37. 2:17If it’s a cell phone, yes, pollsters can (and do) call cell phones, they’ll just
38. 2:20make sure the person is over 18.
39. 2:22When it comes to election polling, they’ll also ask if the person they’re surveying
40. 2:26plans to vote; if you don’t plan to vote, your opinion doesn’t really matter -- at
41. 2:30least at least to the outcome of the election.
42. 2:32Then they’ll ask who they plan to vote for.
43. 2:34It’s important this question comes first, and that the order the candidates are listed
44. 2:39in is switched up.
45. 2:40Asking about policy questions first can make a participant realize they don’t actually
46. 2:44agree with a candidate.
47. 2:46Constantly listing one candidate first can also impact the survey, especially in primaries
48. 2:51where there can be several candidates and it’s hard to keep track of them over the
49. 2:54phone.
50. 2:55All these precautions are things online polls don’t do.
51. 2:59Their samples consist of just the website’s readership, which could skew to one area,
52. 3:03one age group, one political bias, or people who won’t even vote.
53. 3:07News organizations love to say, "our online poll shows ____."
54. 3:11But, their online poll isn't a representative sample, and doesn't actually, "show" anything,
55. 3:16except what the people who come to their website might think.
56. 3:18Sometimes political ads even masquerade as polls, asking questions about a candidate
57. 3:23in a negative way to sway opinion.
58. 3:25Because scientific polls strive to eliminate as many of these confounding factors as possible
59. 3:29they are much more reliable than online public surveys.
60. 3:33If you want to check out how well they perform, the National Council on Public Polls keeps
61. 3:37track of how accurate each poll was after every election.
62. 3:42You can check out their link in the description below.
63. 3:44In 2012 Pew predicted Obama would win 50% to 47%.
64. 3:48The final tally?
65. 3:5051 to 47.
66. 3:51We think we’re pretty special with our precious Democracy but it isn’t uniquely human.
67. 3:55Julian and I explore other species that vote here.
68. 4:13So even though open online polls are essentially meaningless, they promote audience engagement.
69. 4:19Do you think a poll is the best way to do that or is there a smarter way to get people
70. 4:23involved?