Poll Watch: Much Ado about Nothing Much
We're over a year away from the next presidential election, and even the earliest state primaries are more than a half-year away, but apparently it's not too soon to start polling. The primaries and conventions are so distant that it's unlikely that a poll will tell us much even about who will be nominated, let alone about who will win in November of next year. Nonetheless, here is the headline of a news article about a recent poll:
Walker leads nationally in new poll
What does that mean? Does it mean that Governor Scott Walker of Wisconsin is ahead of Hillary Clinton in the poll? Not at all. Here are the first two paragraphs of the article following the headline:
Wisconsin Gov. Scott Walker leads a tight field of candidates for the Republican presidential nomination, according to a new survey from Public Policy Polling. Walker is alone in first place in the poll with 17 percent, followed by former Florida Gov. Jeb Bush at 15 percent, Sen. Marco Rubio (Fla.) at 13 percent, retired neurosurgeon Ben Carson at 12 percent and former Arkansas Gov. Mike Huckabee at 11 percent.
Source: Jonathan Easley, "Walker leads nationally in new poll", The Hill, 6/16/2015
Now, if you know that almost all national public opinion polls have a margin of error (MoE) of at least plus-or-minus three percentage points, you'll immediately spot a problem here: The difference between Walker and his nearest competitor, Jeb Bush, is only two percentage points, which is within the MoE. Similarly, the difference between Bush and the next candidate, Senator Marco Rubio, is also two percentage points. Finally, Rubio differs from the following candidate, Ben Carson, by a single percentage point, and Carson is only one point ahead of Mike Huckabee. Thus, none of the differences between any of the top five candidates and their nearest competitors is statistically significant.
However, the situation is worse than that, though this particular article does not supply the relevant information. For that information, you have to go to the report on the poll put out by Public Policy Polling―see the Source, below. At the end of The Hill article, we're told that 1,129 people were polled for a MoE of 2.9 percentage points, which is about what we would expect. However, as we learn from the polling report itself, the comparisons of the Republican candidates came from a subsample of 429 Republican voters, and so the MoE for these results is actually ±4.4 points. Thus, Walker, Bush, and Rubio are all within the MoE of one another.
What can we learn from this poll and the reporting of it?
- The news media usually report poll results as races with winners and losers even when there is no clear winner. Reporters frequently treat poll results as if they are perfectly precise, despite mentioning the MoE at the end of the article. Whether this is because of a lack of understanding of the MoE, or they ignore it in the interest of making a more dramatic and exciting story, the result is the same.
- At this early date, it's likely that most Republican voters have not paid sufficient attention to the campaign to choose a favorite candidate, and the poll results showing several candidates close together probably reflects that fact.
- Because it's so early in the race, there may be hundreds of national political polls between now and November of next year. Because of the 95% confidence level used to determine the MoE in polling, we can expect 5% of these polls to have results that differ from reality by greater than the MoE. Thus, there are bound to be tens, and perhaps scores, of polls before voting is over that will be outliers. With so many polls being conducted, it's very important to compare poll results: if a given poll shows very different results from others conducted around the same time, it's most likely that the outlier is wrong.
Source: "Walker, Bush, Rubio lead GOP Field Nationally, Clinton Still Dominant", Public Policy Polling, 6/16/2015
Resource: How to Read a Poll
The Case of the Puzzling Statement
A suspect was arrested on a charge of murder. When interrogated by the police, the suspect volunteered to write a statement if he was left alone with a sheet of paper and a pen. The police proceeded to do so, but a lawyer hired by the suspect's wife showed up soon thereafter, demanding to see the suspect in private. Afterward, the suspect refused to write or say anything further on the advice of his attorney. All that was written on the piece of paper was the following:
If the victim was asleep when I arrived but the front door was locked, or the back door was open, then either I killed him or both a light was on in his room near midnight and the murder weapon was a gun, unless neither was he drunk when he died nor was the murder weapon a knife.
Unsurprisingly, the police are baffled by this statement and have very little evidence against the suspect. All that they know for sure is that the victim was not shot but killed with a knife, and that the back door had been left open. If they can't decipher the cryptic statement, the suspect will surely walk. Can you help the police make sense of the suspect's statement?
Assuming that the suspect's statement is true, and that the evidence gathered by the police is correct―that is, that the victim was killed with a knife, not a gun, and the back door was open―can you determine whether or not the suspect killed the victim? If you think you know the answer, click on "Solution", below, to find out if you cracked the case.
New Book: Spinglish
"Spinglish" is a combination of the words "spin" and "English"―in analogy to the more familiar term "Spanglish"―describing English words and phrases used to "spin". Henry Beard and Christopher Cerf have produced a book of that title, subtitled The Definitive Dictionary of Deliberately Deceptive Language. Beard and Cerf are best known for writing humor books, so I'm not sure how serious a work this is. Also, having not yet read it, I don't know whether it adds anything much to William Lutz' doublespeak dictionary, or Steven Poole's Unspeak. Some of the words and definitions I've seen attributed to the book were discussed previously by Lutz or Poole. However, I do think it will be a useful reference even if it just gathers together all of the examples from previous writers in dictionary form together with citations.
- Henry Beard & Christopher Cerf, Spinglish: The Definitive Dictionary of Deliberately Deceptive Language (2015)
- Ron Charles, "‘Spinglish,’ by Henry Beard and Christopher Cerf", The New York Times, 5/23/2015
Update (6/27/2015): There's an excerpt from Spinglish in The Daily Beast consisting of the introduction and all of the entries in the book that fall under the letter "G"―see Source 1, below.
Some of the words are new to me, such as "gallerist", which is an inflated term for "art dealer"; others are familiar, for instance, "gaming" used as a general euphemism for the more specific "gambling"―William Lutz includes it in his 1999 book Doublespeak Defined, see Source 2, below. I wrote about "government relations professional" back when it was in the news as a proposed euphemism for "lobbyist"―see the Resource, below.
Some of the definitions are themselves tendentious, for example, "[t]he hunting or mass slaughter of wild animals" as a definition of "game management". It's true that "game management" is often used as a euphemism for the more specific "hunting", but throwing in "mass slaughter" adds nothing to the definition but emotional bias against hunting as a means of controlling the population of animals. Moreover, if you're going to include "gun grabber" as "[a] term for someone who supports gun control legislation favored by those who don’t", why not include "gun nut" as "a term for someone who opposes gun control legislation favored by those who support it"? There seems to be a degree of spin in what counts as "spinglish".
However, some of these words are not like the others; some of these words just don't belong. For instance, "greenwashing" is not a euphemism, but if anything the opposite, since it's used by environmentalists to condemn businesses that promote their practices as "green" when they're really not. This is a problem with introducing a new word, such as "spinglish" or "unspeak", no matter how clever and amusing it may be: unless such a word is precisely defined, it's hard to know what counts as an instance. The closest that Beard and Cerf come to defining "spinglish" seems to be the following passage from the introduction:
But what precisely is Spinglish? Well, in spite of its polyglot-sounding name, it isn’t some foreign language. It’s just our native tongue, transformed into a sophisticated method of judicious miscommunication through the use of careful word choice and the artful rephrasing and reframing of familiar terms. To put it another way (which, of course, is what Spinglish is designed to do), it all comes down to making me sound better, or you sound worse, or both.
Unfortunately, that isn't a very precise answer. "Careful word choice" and "artful rephrasing" are characteristics of good writing. The only real help in this "definition" is that it characterizes spinglish as "a sophisticated method of judicious miscommunication". In other words, what seems to separate spinglish from good English is that the latter aims at communication whereas the former's goal is obfuscation. However, it's perfectly possible to mislead people without using the euphemisms, doublespeak, loaded language, and weasel words that seem to constitute most of the entries in this book.
"Greenwashing" doesn't seem to meet this vague definition, since it serves as a label for a certain type of spin. If "greenwashing" is spinglish, then would "spinglish" itself count as spinglish?
I hope I don't sound too much like a pedantic killjoy. Beard and Cerf might well reply that their book is intended as amusing and entertaining light reading rather than a scholarly monograph, and my goal is not to spoil anyone's enjoyment of it, but to point that out.
- Henry Beard & Christopher Cerf, "How to Translate ‘Spinglish’", The Daily Beast, 6/23/2015
- William Lutz, Doublespeak Defined: Cut Throught the Bull**** and Get the Point (1999), p. 153
Resource: Doublespeak Dictionary, 9/24/2013
Update (6/28/2015): Here's an up-to-date example of spinglish, doublespeak, unspeak, or just plain barnyard epithet. The phrase is "gender confirmation surgery" for what used to be called "sex change surgery", and more recently "gender reassignment surgery". None of these terms is accurate, since you can't completely change your sex, and what does it mean to say that you "confirm" or "reassign" someone's gender by surgery?
England's Daily Mail has criticized the U.S. State Department for using "gender confirmation surgery" to refer to Iran's practice of forcing homosexuals into such surgery.
The law defines transgender persons as mentally ill, and the government provided transgender persons financial assistance in the form of grants of up to 4.5 million tomans ($1,660) and loans up to 5.5 million tomans ($2,030) to undergo gender-confirmation surgery. In addition, the Ministry of Cooperatives, Labor, and Social Welfare requires health insurers to cover the cost of gender-confirmation surgery. Individuals who underwent gender-confirmation surgery may petition a court for new identity documents with corrected gender data, which the government reportedly provided efficiently and transparently. Human rights activists and NGOs reported that some LGBT persons were advised to undergo gender-confirmation surgery to avoid legal and social consequences due to their gender-identity ambiguity.
- "Country Reports on Human Rights Practices for 2014: Iran", U.S. State Department, Section 6, "Acts of Violence, Discrimination, and Other Abuses Based on Sexual Orientation and Gender Identity"
- Wills Robinson, "State Department refers to Iran's forced sex changes of homosexuals and transgenders as 'gender-confirmation' because their identity was 'ambiguous' beforehand", Daily Mail, 6/27/2015
Solution to the Case of the Puzzling Statement: According to the police, the back door was open, which means that the following is true: the victim was asleep when the suspect arrived with the front door locked or the back door was open. We don't know whether the victim was asleep or the front door was locked, but we don't need to since we know that the back door was open. So, we can conclude everything after "then" in the suspect's statement.
According to the police, the victim was stabbed not shot, so we know that the murder weapon was a knife. We don't know whether he was drunk when he was killed, but we don't need to because we know that it's false that neither was he drunk nor the weapon a knife. Therefore, we can conclude that either the suspect killed the victim or both a light was on in the room near midnight and the murder weapon was a gun.
However, the murder weapon was a knife, not a gun, so it's false that there was both a light in the room near midnight and the weapon was a gun. We don't need to know whether the light was on to know that it's not the case that both are true. Therefore, the suspect killed the victim.
Technical Appendix: For those who know propositional logic, here is a proof that the suspect killed the victim. Let's use the following abbreviations:
A: The victim was asleep when the suspect arrived.
B: The front door was locked.
C: The back door was open.
D: The suspect did it.
E: A light was on in the victim's room near midnight.
F: The murder weapon was a gun.
G: The victim was drunk when he died.
H: The murder weapon was a knife.
Using those abbreviations, together with standard symbols for truth-functional relationships, we can abbreviate the suspect's entire statement as follows:
[(A & B) v C] → [D v (E & F) v ¬(G v H)]
This, of course, is just one way of translating the statement, but if you came up with a different one it should be logically equivalent. In addition, the evidence gathered by the police can be represented as the sentence:
C & ¬F & H
From the statement and evidence, we can use natural deduction techniques to prove that the suspect did it (D):
- (A & B) v C, from the evidence by Simplification (Simp.) and Addition (Add.).
- D v (E & F) v ¬(G v H), from the statement and 1 by Modus Ponens.
- G v H, from the evidence by Simp., and Add.
- D v (E & F), from 2 and 3, by Disjunctive Syllogism (D. S.).
- ¬E v ¬F, from the evidence by Simp. and Add.
- ¬(E & F), from 5 by De Morgan's Theorem.
- D, from 4 and 6, by D. S.
This is the proof that was given in English in the Solution, above. Again, this is just one way to prove D using natural deduction. It's also possible to use a truth-value assignment technique to show that D must be true for the statement and evidence to be true, but I'll leave that as an exercise for the interested reader.
Update (7/1/2015): I have rewritten the statement of the puzzle and solution to remove an ambiguity called to my attention by an alert reader who prefers to remain anonymous.