Wikipedia:Reference desk/Archives/Miscellaneous/2016 October 18

Welcome to the Wikipedia Miscellaneous Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

October 18

Polls

What's with the yuuuge spread among the various political polls this year? Even this close to the election, some show Clinton leading by double digits, while others show a statistical tie or even Trump slightly ahead: http://www.realclearpolitics.com/epolls/latest_polls/president/ If they're scientific polls, you'd expect them to be within about 5-6 points of one another, no? So what's the reason for this spread? 2601:646:8E01:7E0B:8085:B0C0:23E8:D579 (talk) 00:51, 18 October 2016 (UTC)

Apparently yuuuge is a strange pronunciation of "huge".

Bernie does it too. Trump was born in Noo Yawk in the 40s like him. Grew up in Noo Yawk like him. Is Caucasian like him. It's no wonder they both tawk with a strong Noo Yawk accent. This pronunciation of 'huge' is supposedly uncommon in English-speakers as a whole. [1] Sagittarian Milky Way (talk) 17:53, 25 October 2016 (UTC)

There are a number of factors at play:

1) Third party candidates may make a difference in this election. So, whether the poll is just between Hillary and Trump or includes the third party candidates may affect the outcome.

2) This election may change who votes and who doesn't vote. So, whether you poll all people, only registered voters, or only likely voters, you may get a different result, and figuring out who the likely voters are is no easy task.

3) As always, whether you look at popular vote or electoral votes, you may get different numbers.

4) Some of the so-called "scientific polls" may be cherry-picking results to favor their candidate. I'd look at real news orgs that have called elections correctly in the past, versus fake news orgs who tend to favor one party. StuRat (talk) 01:10, 18 October 2016 (UTC)

But how can you tell which news orgs are reliable and which are not? Especially after such a (formerly) respected news org as Reuters got caught blatantly cooking the poll numbers! 2601:646:8E01:7E0B:8085:B0C0:23E8:D579 (talk) 04:34, 18 October 2016 (UTC)

You have to look at their record. Have they reliably predicted the winning candidate in the past ? Of course, this automatically excludes any new polling organizations, as we probably should, until they establish a record. StuRat (talk) 17:46, 22 October 2016 (UTC)

FiveThirtyEight has a reputation as a good site for meta-analysis of polling results, knowing which polls are good and which aren't, and compiling data from the good ones. --Jayron32 01:36, 18 October 2016 (UTC)

• I don't know if the spread is any bigger this year than previously. In 2012, even one week before the election Ipsos (well-respected) gave Obama a 12 point lead while Gallup (also well respected) gave Romney +5. Smurrayinchester 09:42, 18 October 2016 (UTC)

Is that so? 2601:646:8E01:7E0B:F88D:DE34:7772:8E5B (talk) 11:00, 18 October 2016 (UTC)

Polls, part 2

A follow-up on my previous question: Just how does one deliberately skew a scientific poll? (I know precisely one method for doing this -- reclassify one category of respondents and lump it in with the one you want to artificially enlarge, which is what Reuters/Ipsos had done (they lumped in all the "None of the above"s with Clinton supporters) -- but are there others?) 2601:646:8E01:7E0B:F88D:DE34:7772:8E5B (talk) 11:04, 18 October 2016 (UTC)

Opinion_poll#Potential_for_inaccuracy has some ideas. --Jayron32 11:59, 18 October 2016 (UTC)

Weighting methods are mentioned there as a remedy, but they can also in concept be intentionally altered to mislead. AAPOR has additional info on poll weighting here [2]. Here's a recent story [3] about how the weighting proceedure of one poll has influenced its results. SemanticMantis (talk) 14:26, 18 October 2016 (UTC)

• In response to Stu, Fox News has consistently been publishing polls with Hillary ahead, 5 pts today, and the LA Times has Trump up by two. But the only appropriate answer is that polls are not scientific, since they, (1) are not replicable, (2) are not peer-reviewed, (3) have "weighting" that is hidden (similar to climate science, where data is manipulated by a model to fit a prediction) and don't represent any actual entity. What matters is the results in the Electoral College and the House if it gets that far. All else is pretense. Dewey Defeats Truman. μηδείς (talk) 00:09, 19 October 2016 (UTC)

Note that this election is a bit different, because many Republicans don't support Trump, who really doesn't seem to be a Republican himself. And then there's his extremely "non-presidential behavior" and his closeness to Putin. (Makes me wonder if Trump is the lead in a false flag operation, working for the Democrats to discredit Republicans.) So, polling orgs that historically may have laid their credibility on the line to fake results to give Republicans an edge may not be so tempted this time around. StuRat (talk) 14:47, 27 October 2016 (UTC)

Where's the evidence that climate science is manipulated to fit a prediction? The prediction is common sense too: raise Earth's CO₂ levels from 280 to 400 and temperatures will rise. Now the onus is on you to tell by what mechanism more greenhouse gases don't cause more greenhouse effect. Sagittarian Milky Way (talk) 01:38, 19 October 2016 (UTC)

I said like climate science where the data is manipulated to fit a model, not simply "like climate science", of which I have read plenty which is purely empirical. And Fox now has Hillary up by six, not five. How can you have any pudding if you won't eat your meat? μηδείς (talk) 03:44, 19 October 2016 (UTC)

Either show evidence of data manipulation in climate science or retract that rather serious accusation. Fgf10 (talk) 06:41, 19 October 2016 (UTC)

Thus spake The Nanny. ←Baseball Bugs ^{What's up, Doc?} carrots→ 10:16, 19 October 2016 (UTC)

Extraordinary claims require extraordinary evidence. Do you have something useful to add? If not, don't post. Fgf10 (talk) 10:45, 19 October 2016 (UTC)

Take your own advice. Just below is the right way to challenge someone's statement. Yours is the wrong way. If you can't do it the right way, then you have nothing useful to add, and should not post. ←Baseball Bugs ^{What's up, Doc?} carrots→ 11:13, 19 October 2016 (UTC)

Fgf10's claims are not extraordinary. It's and is never necessary to provide sources to challenge unsourced claims on the RD, whatever StuRat may and his ilk may like to claim. In fact, in a number of cases it may not even be possible to provide reliable sources to challenge a claim simply because there isn't a wikipedia fact checker where everyone random statement by some random person here is challenged. (Although it's likely reliable sources could be found for this case.) This is actually the way world normally works, again whatever StuRat and his ilk may like to claim. If you're going to make extraordinary claims you should at least be willing to provide evidence when challenged. Yes I know a certain presidential candidate seems to think if you keep repeating stuff it becomes true even if everyone says it's not, often with good evidence, but that doesn't mean we should follow them. Of course μηδείς is the one always moaning about how this is supposed to be an RD, wanting to delete everything because it wasn't sufficiently on topic for them etc, but this is hardly the first time this has happened. Nil Einne (talk) 02:11, 21 October 2016 (UTC)

You need to re-read what I said. Smurrayinchester's attitude in his response is the right attitude: polite and factual. Fgf's is the wrong one: confrontational and devoid of any facts. ←Baseball Bugs ^{What's up, Doc?} carrots→ 18:26, 25 October 2016 (UTC)

Points 1 and 3 are not correct. Opinion polls are replicable - multiple companies can perform a poll using similar methods and compare results - and reputable polling companies will release the raw data and their weighting methods (this is how people like Nate Silver can build polls of polls). In the UK, this is overseen by the British Polling Council - I'm sure there's a similar entity in the US. You can see, for example, Fox News's latest poll, with all the raw data. Smurrayinchester 10:58, 19 October 2016 (UTC)

(The LA Times data is here, and you can read about their methodology and why it differs here - long story short, their system boosts candidates with very enthusiastic support, which helped Trump through much of the cycle). Smurrayinchester 12:07, 19 October 2016 (UTC)

The US equivalent to BPC is the National Council on Public Polls [4], though "overseen" is rather strong language. Both BPC and NCPP set standards and expectations for polling but they have no power to enforce them. At best all they can do is certify whether or not a pollster is compliant with their recommendations. Many pollsters, including some high profile ones, are not compliant though. Dragons flight (talk) 11:35, 19 October 2016 (UTC)

A few ways to bias a poll:

1) Ask questions in a form other than what they will see on the ballot. So, instead of asking if they will vote for "Hillary Clinton" or "Donald Trump", you ask if they support the "Democrat" or "Republican", the "liberal" or the "conservative", or, if you want to get really blatant about it, the "Washington insider" versus the "successful businessman".

2) Choose a biased sample. So, ask the question at a National Rifle Association rally in Montana or at a black church in an inner city.

3) Cherry-picking: Do a large number of small polls, ensuring a large margin of error, then toss out the ones that don't give the desired result, and keep the rest. This can be done indirectly, by hiring companies to do you polling. Those that provide the numbers you want, you hire again, the rest you fire. Pretty soon the survivors find ways to give you whatever results you want.

4) Toss out individual polling results that run counter to your desired result. So, if they vote the "wrong way", you disqualify them if they listed a middle initial instead of the full name, as requested, while you ignore such trivial mistakes on those who voted the "right way".

5) There can be some subtle mathematical methods. For example, if two equal polling groups give you 40.6 and 41.6 percent, you can round each to 41 and 42, then average the two, to get 41.5, then round that up to 42, or you can average them both first to get 41.1, then round that down to 41. If you do these rounding tricks at the district level, state level, regional level, and national level, you could have several percentage points difference. StuRat (talk) 17:56, 22 October 2016 (UTC)

The fundamental problem is Hume's problem of induction, that you can't say that the future will be like the past. I'll use the US election 2012 as an example: Republicans assumed that polling well among registered-independent voters meant that they would win. What they didn't grasp (or, perhaps, want to admit) was that the Bush administration had made them so toxic that even their own voters didn't want to admit that they were Republicans. So "independents" (actually moderate Republicans) voted Republican - but it didn't matter, those were the voters that would have been registered Republicans ten years earlier. Romney lost by a lot, while winning independent voters.
Similarly, this US election there is no black person on the ballot, while there is a woman on the ballot. It's reasonable to suspect that this will depress black turnout and raise female turnout - but by how much in each direction? The answer could make a lot of difference. More early votes are coming in than last cycle - but are those people who didn't vote last time, or has one party's committed voters disproportionately decided this year to vote by post?
More loosely, there's always some weaker, less scientific indicator you can point to as proxy that you're winning: "So many people turn up to our rallies!" (we have a candidate that the party base loves but nobody else does) "Sure, we're behind right now, but people say in polls that they're unhappy with the direction of the country - we just need a few more weeks to get our message across!" (They may not like where they are now, but they still would rather that than the crazy orange guy.) "The rule of the undecideds says undecideds break 70% for the challenger - with many undecided voters, we should win even though the polls say we're behind!" (But why is that going to be true this time?) Et cetera. Blythwood (talk) 04:33, 23 October 2016 (UTC)

By the way, if anyone doubts that anyone could ever be as shameless as some of StuRat's examples, guess again. I've lived in an area where the council launched an online questionnaire which had some insane questions. I'll fictionalise it to preserve anonymity, but imagine something like the question "Do you support the council taking action to find funds to restore these old buildings?" as code for meaning "Do you support the council selling off land to a housing developer?" It then wrote a report discounting replies after local pressure groups found out about the questionnaire and posted details of it on a Facebook page, since all the resulting replies were clearly the result of "misinformation". I want to stress that reputable polling firms do not do this kind of thing, but if you are a private organisation running an "informal" consultation, you can get away with much more. Blythwood (talk) 04:49, 23 October 2016 (UTC)

Another common way in which surveys are biased - often accidentally - is through the question order. For example, if you wanted to get a strong result for Trump, you might first ask a series of questions relating to e-mail hacking, perceptions of Washington insiders, etc. Even asking which party they normally support first could potentially bias the poll - for example, if someone usually supports the Republicans but is considering a vote for Clinton or Johnson, this prior identification with the party may make them more likely to give the name of the party's candidate. Warofdreams talk 22:49, 24 October 2016 (UTC)

Reminds me of this scene from Yes Prime Minister: https://www.youtube.com/watch?v=G0ZZJXw4MTA. PrimeHunter (talk) 00:03, 27 October 2016 (UTC)

The most accurate predictions based on polling information can be found here Count Iblis (talk) 00:41, 25 October 2016 (UTC).

On this topic, a comment I can remember from a US political commentator, about recent election rallies headlined by Obama (paraphrased since I can't find it): "You can tell who's the professional and who isn't. Trump says 'Vote for me.' Obama says 'Register to vote.' And then he reads out the website address." Knowing how to help your supporters to vote counts, especially if your base is poorer voters. Blythwood (talk) 12:24, 25 October 2016 (UTC)

I was surprised to read a few days ago that Donald Trump used to be a Democrat. 86.128.234.239 (talk) 14:49, 25 October 2016 (UTC)

For a non-politician, he sure has glommed on to the idea of changing his position to get votes from whoever he is courting at the time. For example, he says racists things when he speaks to racists, and avoids them otherwise. I'm not sure this strategy works anymore, though, due to all the tiny recording devices everyone has now. StuRat (talk) 15:30, 25 October 2016 (UTC)

Well his dad was arrested at a Ku Klux Klan rally after it went violent and they both settled one of the biggest lawsuits brought by the Justice Department ("Ministry") for housing discrimination against black people in US history [5] so that might be a slight clue as to his true party. Being a Democrat was probably the hip thing in New York social circles so.. June 1967-July 1987: unknown July 1987-October 1999: Registered Republican October 1999-August 2001: Registered Independent (he was running for President) August 2001-September 2009: Registered Democrat September 2009-December 2011: Registered Republican December 2011-April 2012: [No party] (independent) April 2012-now: Registered Republican. Cliffs Notes version: R>I>D>R>I>R. Source for the KKK thing: Trump confirming with New York Times that dad lived on Devonshire, then denying the street when arrest is brought up, then denying the US Census, then denying it happened, then saying he was never charged 7-8 times, then denying it happened. Sagittarian Milky Way (talk) 17:53, 25 October 2016 (UTC)

FYI, the Republican Party being favored by racist White southerners is historically a fairly recent development. See Solid South. The White Racists started to transfer their allegiances to the Republican party in 1964 following the Candidacy of Barry Goldwater and his refusal to endorse Civil Rights legislation, a move that has been noted as the sine qua non of the modern association of the Republican party with White Supremacists (which had previously been universally Democratic prior to then, per the Republicans being the "Party of Lincoln"). Republican strategist Avik Roy explains as such here. However, Goldwater only started the transition, he still lost that election in a landslide to Lyndon Johnson, see United States presidential election, 1964, however you can see most of the Deep South swung Republican, most for the first time in a century. However, most of those states went to George Wallace in 1968, a Southern Democrat who ran as an independent. It wasn't until 1972's Southern Strategy that the Republican party began to directly court that vote and Nixon won in a landslide mostly on the back of the abandonment of the rest of the South from the Democratic party. This holds, however, only in presidential elections. In local elections, the South remained staunchly Democratic (in the pre-1960's Southern Democrat tradition) until well into the 1990s. It was the Contract with America, which was basically the "Southern Strategy" applied to the House of Representatives elections. So really, it wasn't until 1994 that most White Supremacists in the south completely abandoned the Democratic Party and joined the Republican Party in earnest, though the process started some decades earlier. It isn't like the Democrats have always stood for racial justice and the Republicans have always been the party of White Supremacists. Short-short version: Prior to the early 1990s, there were lots of racists who voted strictly Democratic, and in many parts of the country, it was not an oddity, it would be expected that the KKK and others would do so. You'll note that noted KKK leader David Duke (see David Duke#Early campaigns) ran for elections as a Democrat as late as 1988. --Jayron32 18:09, 25 October 2016 (UTC)

True but that's only relevant to what party a white supremacist's real preference would be then, not what it'd be after 2000. Donald was a Democrat till a year after birtherism got big but didn't mention birthism till 2011 when he became their poster child, then Obama showed his long form, then Trump offered Obama's charity $5 mil to show his college/passport applications as if he thought the long form was fake, then he finally said he's a citizen 4 years later (right before election) without any extra evidence which suggests Trump's often a poser. Evidence suggesting a racist real Donald is frequent enough: 1927, 50s, 50s (all circumstantial), 1973-1975, 1978, 1979, 1983, 2015, 2015, 2015, 2015, 2015, 201... ,2016, 2016, 2016, 2016, 2016.... I guess housing discrimination alone wouldn't mean Trump believed inherent white superiority (he could've just thought it'd increase profit) but then he got what's one of the biggest US housing discrimination lawsuits even now and didn't stop. I doubt his life plan in '75 was "fake racism by violating housing law as long as possible so I have helpful demagogue cred when the alt-right happens so I can run for President in 2016. I will get lucky and not get in trouble. Also grope women, that will totally not hurt me. And let the recessions which are needed to make racism cool again make me go bankrupt 5 times even though business acumen will be a big point of my campaign. Praise a likely opponent, that totally fits in with my plan." Sagittarian Milky Way (talk) 23:49, 25 October 2016 (UTC)

"Town hall" definition

A link in your previous post is captioned

GOP presidential candidate Jeb Bush attack [sic] Donald Trump's record as a conservative during a town hall in Merrimack, N.H., on Aug. 19, 2015.

The phrase "town hall" appears to be what we in Britain would call a "hustings" (a Brexit by - election has just been called, so the hustings there will be interesting). The town hall is the building where the local council meets and where its administrative staff have their offices. How and when did the American phrase get its meaning, and what do Americans call the headquarters of the municipality? 86.128.234.239 (talk) 09:51, 26 October 2016 (UTC)

(US) In a town, the administrative building is called the "town hall", and in a city, the "city hall". Since these type of events originally were held in the town hall, they took on that name (called a "town hall meeting" initially). Town council or city council meetings are supposed to be open to the public, so those rooms have seating for an audience. StuRat (talk) 13:02, 26 October 2016 (UTC)

What area like this has the highest population?

It's in the developed world, possible to drive to from the biggest city in the nation but if you try in a 2WD sedan you have at most a 50/50 chance of succeeding.

Inspired by a dream: An Australian tour guide drove us to the start of a road at the edge of the suburbs. He said

• that's the only way for the town of 1,000 at the other end to reach the national road network

• there's not so much as a gas station, ranch, or mine in between

• the road quality's so bad a 2WD sedan has only a 50% chance of finishing it

• if you try it in a regular sedan you might die

and

• the National Coalition kept refusing their requests to make the road easier and this is why 28 million Australians are now run by Labor (implying it was so close that pissing off a village changed the outcome)

We got out and walked, it looked like they just plopped a several inch layer of asphalt on the nature, made the top smooth and left the edge exactly how it plopped on the ground. The smooth asphalt shrunk from 2 lanes to c. 8 feet wide and went from flat to hard for us to walk up. Beyond where we stopped it continued going straight up a 40° slope for what looked like a mile. I asked if it's like this the whole way and he said pretty much. I asked how long it is and he said a tenth of the width of the continent. Sagittarian Milky Way (talk) 20:04, 18 October 2016 (UTC)

You're asking for references to interpret your dream where there was an 8 foot wide road that was 1/10th the width of the continent of Australia? I'm not even sure where to begin researching such a nonsensical request... --Jayron32 20:09, 18 October 2016 (UTC)

I'm just asking what's the largest population region of the developed world that needs a 4WD/off road vehicle to have a good chance of reaching from the "regular" road network. Where like a Corolla or Civic won't cut it (unless you like being towed out of mud or something every other time). Sagittarian Milky Way (talk) 20:20, 18 October 2016 (UTC)

See the page guidelines, and find a web forum. μηδείς (talk) 21:05, 18 October 2016 (UTC)

SMW is clearly asking about a location meeting certain criteria, and any interested responder can supply locations that may meet said criteria, along with references. Nothing in our purview says OP cannot be inspired by a dream to look for certain facts about the world. All participation here remains voluntary, and nobody is mandated to respond with anything. SemanticMantis (talk) 21:36, 18 October 2016 (UTC)

Do you cut and paste that "anybody who doesn't want to" nonsense from somewhere, SM? Unless you are going to post a link or a ref, then you don't have to unhat this "question", do you? μηδείς (talk) 00:03, 19 October 2016 (UTC)

You can look at Peninsula Developmental Road, or Cape York Peninsula for some out of the way places in Australia with not-the-best roads. Another road with 800 km of no-fuel is Tanami Road, but there are other ways to go to Halls Creek, Western Australia. Graeme Bartlett (talk) 21:14, 18 October 2016 (UTC)

• According to our article, Iquitos, Peru is "widely regarded as the largest inland city that is inaccessible by road". Cities like that in the Amazon Rainforest are probably your best bet for a large city with only dangerous road access. Warzones are probably another - millions of Syrians live in besieged areas only accessible by roads controlled by hostile forces (whether ISIS or Assad). Smurrayinchester 08:23, 19 October 2016 (UTC)

(Until 2010, the entire Russian Far East (population 6.6 million) would have fit your definition, incidentally. Now the R297 highway (Russia) is open, it's possible to drive from Moscow to Vladivostok without having to traverse the perilous Zilov Gap.) Smurrayinchester 08:46, 19 October 2016 (UTC)

Although you can drive to Vladivostok year round, getting to Yakutsk and beyond is tricky, as you have to cross the Lena River, either by driving across the frozen water in winter, or getting a ferry in summer - in between, it's not accessible in a regular vehicle, so it might be the best answer to the question. If not, perhaps Norilsk, which isn't on the Russian road network, but has various accounts of people reaching it overland, presumably by driving up the Yenisei River when it is frozen over. Warofdreams talk 16:10, 19 October 2016 (UTC)

• This list here has a mix of remote island cities and mainland cities. From that list, it looks like La Rinconada is particularly difficult to get to. There are also several settlements in Northern Canada that are treacherous to reach by road; it's actually easier to reach them in the winter where ice roads can be made; during the summer thaw it's basically a roadless swamp. Tuktoyaktuk is perhaps one of the more famous such settlements. I think the entire mainland portion of Nunavut is also basically unreachable from the rest of Canada excepting a few ice roads and minor paths over which many vehicles may find difficult to traverse. Following that thread, I did find the Capital of Nunavit, Iqaluit, states "... like the rest of Nunavut, has no road, rail, or even ship connections for part of the year to the rest of Canada." Of course, Iqaluit is on an island, so that doesn't meet the requirements. The largest settlement on Mainland Nunavut is Rankin Inlet. The largest inland city (thus requires road or rail access) is Baker Lake, Nunavut, which MAY be a good call for the most remote inland settlement of 1000 people in the world; looking at Google Maps, it is at LEAST 700 miles from Flin Flon to Baker Lake, and 600 miles to Yellowknife, the two nearest cities "as the crow flies" which are connected by improved road to Canada's national road network. I suspect there are many similarly sized settlements in Russia equally as far from improved roads, but that's the most remote one I could find in North America. And Baker Lake is about 1800 people. --Jayron32 17:58, 19 October 2016 (UTC)

When considering Canada, it's a good idea to look at which areas are served by the Canadian National Railway, since it, and it's adjacent service roads, are often the only access by land to northern Canada. They've closed some of their more northern routes (they now extend further south into the US than they do north, into Canada), and those service roads soon fell into disrepair, so that's one way to search for such isolated communities. Look at the route map in that article to get started. StuRat (talk) 17:29, 22 October 2016 (UTC)

• • La Rinconada is easy to get to. La Rinconada, Peru not so much. jnestorius^(talk) 12:22, 21 October 2016 (UTC)