Causality 101

(In a statistical training we just attended on pooled and panel data analysis, our trainer Dr. Dennis Mapa, a statistician-economist reviewed John Stuart Mill's necessary conditions to establish causality or causal argument. Sans the mathematical proofs that sometimes rob us of the intuition , he so clearly summarized that TO ESTABLISH CAUSALITY, one has to establish that (1) the cause preceded the effect; (2) the cause was related to the effect; and (3) we can find no plausible alternative explanation for the effect other than the cause. He added that TO BE ABLE TO USE A RESULT FOR FORECASTING, one has to establish a fourth condition: (4) Past values of your outcome variable does not affect current values of your policy variable. And finally, TO BE ABLE TO ESTABLISH A POLICY EFFECT, one has to establish that the beta coefficient for that policy variable is not changing through time. 

While reality in empirical research is limited by resources--time, budget, willing subjects-- so that a "perfect data" is, most of the time never available and difficult to produce, the discussion reminded me to at least try my best to always have these conditions in mind from conceptualization to data gathering to data analyses to report-writing. Anyway, below is an exercise I found in the net on Mill's methods/conditions on causal arguments that drives the points very simply.)

 Exercise 20. Answer the following questions, and explain your answers.

i. Are causal arguments required to explain how the purported cause causes the effect?
ii. Are causal arguments required to show that the purported cause is the only thing that ever causes the effect?
iii. Are causal arguments required to show that the purported cause absolutely always causes the effect?
iv. Can a causal argument succeed without showing that the purported cause is at least sometimes correlated with the effect?
v. Can a causal argument be based on a "cause" that hasn't been previously proved to exist?
vi. Can a causal argument be based only on the fact that the "cause" is always absent whenever the effect is absent?
vii. Can a causal argument work if something else is just as strongly correlated with the effect as the purported cause?
viii. Can a causal argument be based on anecdotal evidence?
ix. Does the "cause" have to have a one-to-one correspondence with the effect?
x. Is "it's the only thing I can think of" ever a good causal argument?

Exercise 21. Answer the following questions, and explain your answers.

i. Is it possible to make a successful causal argument without any evidence of a correlation between the effect and the thing that’s supposed to be causing it?
ii. Is it possible to make a successful causal argument without providing an explanation of how the purported cause might cause the effect?
iii. Is it possible to make a successful causal argument without providing any evidence that the purported cause even exists?
iv. Is it possible to make a successful causal argument if it is proven that the purported cause sometimes happens without being followed by the effect?
v. Is it possible to make a successful causal argument based on a single incident?
vi. Is it possible to make a successful causal argument if it is proven that the effect sometimes happens even though the purported cause is not present?
vii. Is it possible to make a successful causal argument based only on the fact that no-one else can think of anything that could cause the effect?

Exercise answers:

20.
i. No.
ii. Nope.
iii. Nuh-ah!
iv. Well, no.
v. Good god, no! (oops, he can say it better next time-che)
vi. No way.
vii. No. Are you kidding?
viii. Noooooooo.
ix. Ha! No.
x. Don’t make me laugh!
(All answers are "no.")

21.
i. No.
ii. Yes.
iii. No.
iv. Yes.
v. No.
vi. Yes.
vii. No. 

(SOURCE:  Martin C. Young, 2009 see full article at http://www.madwizard.com/ctl_causal.htm)