I. Causation and Correlation

The most common way to argue for a general causal claim is by establishing a correlation. Notice how well this fits with the proposed interpretation of general causal claims.

But there can be correlations without causation. There are at least four ways this can happen:

 

a) factors reversed
b) common cause
c) accident
d) “tag along” factor 

 

We can apply these - not very plausibly - to the smoking example.

 

a) factors reversed - having lung cancer causes smoking. The timing is wrong for this.
b) common cause - some other factor, say, a genetic factor, both makes people finding tobacco appealing and causes the disease. But smoking itself makes no difference. On this view, smoking is not a necessary element of the set of sufficient conditions. But it is a causal consequence of the factors that are.
c) accident - it’s just a coincidence, with no causal connection at all. The numbers are too big to make this at all plausible.

d) “tag along” factor - suppose standing in doorways, or lighting matches, or something else that tags along with smoking is the cause. This is similar to (c), in that the factors are correlated without being causally connected at all. But, in this case, it is not a coincidence. It still is a kind of “causal accident” that the two things are correlated. This is highly implausible in this case. This is in some ways like a common cause, but these factors are not causes of smoking. This is something like an accident, in that the two factors are not causally related. But it differs in that it isn’t a statistical anomaly that is apt to be removed by further testing. It’s not just a coincidence. There’s a reason smokers have more lung cancer.

 

Think of this in terms of best explanations: is the causal connection the best explanation of the correlation, or is one of these others also a good explanation.

 

An example: p. 292, #2a

Correlation: Being a philosophy a major is positively correlated with getting a high score on the Law Boards among students who take the Law Boards.

Causal claim: same factors.
Alternative formulation of causal claim: Studying philosophy causes good performance on the Law Boards among students.

Could this be a case of correlation without causation? If so, it must be one of the four types described:

i) Can't be factors reversed. ii) Not likely to be an accident. iii) Possibly, there is a common cause. Interest in a certain kind of thinking, or talent for it, might incline people to like phil. and so take it. The course of study may not increase scores. (How would you find out?) iv) Hard to think of anything in this case.

 

#2b - salaries of college grads

Correlation: Being a college graduate is positively correlated with having a high salary among 35 year olds.

Causal claim: Being a college graduate causes getting a high salary among 35 year olds.

There could be common causes here - intelligence, initiative, family background.

 

 

II. Causal Arguments

Example 1: Country Music Linked To Suicide

A recent report says that country music might trigger suicide. Cities in which more country music is played on the radio have significantly higher suicide rates, according to a study that compared the suicide rates with country's radio market share in 49 metropolitan areas, including Nashville. "Country music reinforces thoughts of hopelessness among potential suicide victims - that things are not going to get better and you just have to adjust to these miserable conditions," said a co-author of the study.

 

Causal Argument
1. Listening to country music (C) is positively correlated with committing suicide (S) among people (P).
2. If C is positively correlated with S in P, then either the causal factors are reversed in this correlation statement, or this correlation is the result of a common cause, or this correlation is an accident, or C causes S in P.
3. The causal factors are not reversed.
4. The correlation is not the result of a common cause.
5. The correlation is not an accident.
6. C causes S in P. (1) - (5)

 

Two comments on the pattern of causal arguments:

a) (5) is taken to say that it is not an accident of either of the two kinds mentioned above - coincidence, tag-along factor. [This is a little different than in the textbook.]

b) You could replace all of (2)-(5) with the premise:

 

            C causes S in P is the best explanation of (1).

 

Some questions about this argument:

1. The researchers did not directly measure the suicide rate among country music listeners. So, premise (1) of this argument is inferred from the correlation they actually did measure. What is that correlation?

A couple of answers are fine:

Being a city in which a lot of country music is played on the radio is pos. cor. with having a high suicide rate among cities. [Cities in which country music is popular tend to have higher suicide rates.]

Living in a city in which a lot of country music is played on the radio is pos. cor. with living in a city in which there is a high suicide rate among people. [People who live in cities in which country music is popular tend to live in cities with high suicide rates.] 

 

2. It could be that the correlation statement just given is true but (1) of the causal argument is false. How?

It could be that there is a high suicide rate in these cities, but that the people who commit suicide do not themselves listen to the music. [The country music fans induce suicide in others.]

Another acceptable answer: the correlation could hold in these cities but in other places listeners to country music don't commit suicide, so that overall (1) is false.

 

3. Each of the following statements may raise an objection to this argument. For each statement, say which premise it is best taken to be an objection to and discuss it.

A. Country music appeals mainly to people who are already depressed and suicidal.

This suggests that being depressed is a common cause of C and S. So, it criticizes (4).

 

B. The depressing character of country music has nothing to do with it. It's just that the music is so bad that people who listen to it are driven to suicide.

This gives an alternative way in which C might cause S. It does not attack any premise.

 

C. Country music is popular mainly in impoverished areas of the country where the suicide rate is higher than in other areas.

This suggests that it is an accidental accompaniment of C - living in impoverished areas - that leads to S. So, it attacks (5).

 

4. Which of the following are consistent with the conclusion?

D. The vast majority of people who listen to country music do not commit suicide.
E.. Most people who commit suicide do not listen to country music.
F. In some cases, listening to country music has improved the spirits of suicidal people and thus helped prevent them from committing suicide.
G. No one has ever committed suicide simply because of listening to country music. The explanation is always more complicated than that.
H. A lot of country music is uplifting and optimistic. It isn't all depressing and filled with hopelessness.

Ans.: All are consistent with the conclusion.

 

Example 2: Tennis and Confidence (Ex. 12, p. 311) [Not Done in Class]

The causal claim here is: Being confident (CON) causes being able to come back (CB) among tennis players.

This is based on the correlation: CON is positively correlated with CB among tennis players.

In the discussion that follows we assume that the male - female comparisons are not important. We also assume that CB is what is being studied. We could have said that CB is supposed to test some more general trait - “how one reacts to adversity.” We’ll use the following abbreviations:

Sample:
R: top ranked tennis players studied
U: USTA players studied
S: R+U

Target pop: All tennis players (TP)

Measured properties:
CB: winning a match after losing the first set (coming back)
T: being top-ranked (i.e., being in R)

Target property:
CON: being confident

The argument for the correlation is as follows:

Background: 20,000 USTA matches of players of all ages and abilities were analyzed to see how often players won the match after losing the first set. A similar study was done on top players.

1. Results: 12% of the Us were CB;
37% of the Rs were CB. (EP)
2. Measured correlation: T is positively correlated with CB in S. (1)
3. Accuracy premise: If T is positively correlated with CB in S, then CON is positively correlated with CB in S. (IP)
4. Target correlation in sample: CON is positively correlated with CB in S. (2), (3)
5. Representativeness premise: If CON is positively correlated with CB in S, then CON is positively correlated with CB among all tennis players (TP). (IP)
6. Conclusion: CON is positively correlated with CB in TP. (4), (5)

 

The causal argument, starting from (6), can be formulated by simply plugging the relevant factors into the standard pattern.

 

Evaluation: The correlation argument assumes that being top ranked and having confidence really are correlated. We have no evidence for that given here. We must rely on the author’s testimony. There’s room for doubt, but let’s concede that for now.

 

There are some problems with the causal argument. The general issue is whether there is something else that distinguishes Ts from Us that could account for the difference. That is, the question concerns what the answer to the two questions discussed in the text, p. 303, are when applied to this argument. There are several ways in which top-ranked players are unlike other players, and these differences could also explain their ability to come back. Only one possibility will be described here: top ranked players are better able to adapt their strategy to different players. Lesser players may have to stick to their standard game. So, Ts may come back because of strategy changes, not CON. This would make the correlation the result of a “tag-along factor.” Being confident tags along with adaptiveness, which is the real cause. A good way to think about this: suppose you had two players with equal ability to CB. And you added CON to one and adaptiveness to the other. (Imagine CON injections.) Which would then improve with respect to CB?

 

V. An Essay on Causation

"Clear Thinking About Guns," essay by Charley Reese, D&C, Nov. 27, 1993, p. 326 in text.

Reese takes a generally favorable stand toward logic and critical thinking. He attempts to apply what he thinks he’s learned about this to a variety of topics, especially the proposition that guns cause crime. By looking at what he says about this, and a few of his other points, we can get a better appreciation of some of the central issues associated with general causal statements.

 

A. Reese’s Main Argument

1. If A causes B, then A is both nec. and suf. for B.
2. Guns aren't nec. for crime.
3. Guns aren't suf. for crime.
4. Guns don't cause crime.

On (1):

Smoking cigarettes is neither nec. nor suff. for lung cancer.
Rain is not nec. for wet sidewalks (nor suff, given overhanging trees).
Drunk driving is neither nec. nor suff. for accidents.

But the causal statements associated with these are all true - cigarette smoking does cause lung cancer, etc. So, premise (1) is preposterous. This argument is very weak.

 

B. Is there 1 cause of crime? 

Reese says that opponents of guns are saying that there is one cause of crime. He’s pinning on them the idea that guns are the sole cause of crime. This is absurd - gun control advocates don’t say this. It exemplifies a common strategy in public debates. There’s a problem concerning something, X. Someone advocates doing Y to deal with the problem. Critics of doing Y accuse them of mistakenly thinking that whatever Y will correct is the sole cause of the problem and thus of falsely thinking that Y will completely solve the problem. In this case, Reese seems to be accusing gun control advocates of thinking that gun control will completely resolve crime problems. That’s absurd, but it’s not what they think.

 

What’s particularly ironic is that Reese’s claim about there being multiple causes of crime is inconsistent with (1) of his argument. Suppose that there are two causes, A and B. Then, by (1) each is suff. But then neither is nec. So, there cannot be two. Of course, (1) is what's wrong here

 

C. What’s the Real Issue?

Reese asserts that people who think that guns cause crime are making some big mistake because they are attributing evil qualities to inanimate things. But what is the real issue? The issue isn't whether guns are evil. It's whether having them so readily available contributes to crime (and to evil things people do.) No attribution of moral qualities to inanimate things. Nothing mysterious. So, the causal statement of real interest is something like this:

 

            Having a gun causes the commission of crimes among people.

 

Or perhaps:

 

Having a policy that allows widespread possession of guns causes a high crime rate among societies.

 

It should be clear that we can use correlation statements to attempt to establish these causal claims. When Reese says that a "firearm can facilitate the commission of a crime" he pretty well concedes the main point.