Causation as More than Sequence

Earlier this week I wrote of sequence and causation. If a city experiences both an increase in the crime rate and an increase in unemployment, what can we infer? If you are of one political persuasion, you will surely infer that the criminals drove businesses out of town, leading to an increase in unemployment. If you are of another political persuasion, you’ll as readily declare that the increase in unemployment made people desperate for money, or angry, or both, and led to the higher level of crime.

But that is polemics, not fact finding. Perhaps the truth of the matter is a bit of both, perhaps it is neither. The correlation doesn’t tell you.

Here is an anti-Humean corollary. Even if you have a temporal sequence you can’t securely infer causation. It is true that in economics especially a reliable forecasting of event A by event B, is sometimes called “Granger causality,”  after Clive Granger. But Granger causality isn’t old-fashioned, one-pool-ball-hits-another, no-adjective-required causality.

To see why not, consider two distinct radar systems, one better at longer-range detection than the other. The superior radar system will by definition detect an oncoming airplane before the inferior system will. Thus, there will be a reliable relationship of Granger causality between the detection of a particular blip on the better system and its appearance on the second one a minute later.

But the better radar isn’t causing anything to happen on its inferior cousin. The oncoming airplane is the physical cause of the blips, and the mechanics of the two radar systems are the physical causes of the time lag. Granger causality is not causality sans phrase and we’re still looking for something more.

Or, let’s go back for a moment to the unemployment/crime rate conundrum. The rise of both  might simply have come about because the city in question had a baby boom eighteen years before. The young people are now coming out of high school all at once and, we may suppose, they do so more rapidly than the area businesses can accommodate, so the unemployment rate increases. Also, the crime rate increases, not as a consequence of the unemployment but simply because the demographic bulge has reached the highest crime-rate years, and has passed beyond the age at which indiscretions can be kept hidden in the juvenile system. Either of these two increases can come first, it doesn’t really matter which is observed to start first; there will remain no good reason to believe in a cause-effect hypothesis between crime and unemployment in either direction.

So, what is a good case for the proposition that A causes B? This is a fraught philosophy-of-science question. We can cut through it, I submit, by working from analogy with our discussion of numbers in some earlier posts.  

There are lots of different things called “numbers.” We can understand their relationship intuitively if we accept the fact that there are certain “core” applications of the concept, and other applications that we come to make through later extension, modification, or metaphor. The core of all concepts of number is found in the act of counting. The core of all concepts of causality is found in something even more basic: human action itself.

We learn from very early on that we can move objects by pushing and pulling them. In the act of plucking an apple off a branch, it is impossible to regard the two facts involved [that you have just tugged at the apple, and that the apple is now detached from its branch] as discrete points, connected only by regular succession. No: the pull produced the detachment.  You, the puller, took that apple! This is the primitive notion of cause, and in order to retain their value other notions of cause must be in a position to show their genealogical connection therewith.

Look at the examples we used earlier. The contact between two pool balls on a table causes one to stop and the other to move; the presence of an airplane within range causes the appearance of blips on a radar screen; the increase of a city’s population can increase its crime rate. The first two of those examples involve contraptions of human manufacture. One is simple the other complex, but both are clearly enough seen as cause in an extension of the primal apple-picking sense. They are extensions of human agency.

The third example is a bit of an extension: the rise of the number of humans of a certain age in a given crowded space isn’t usually the consequence of any one’s immediate intentions. Still, the genealogy isn’t complicated. As humans who act in a variety of ways ourselves we recognize that when there are more of us in a small space relations become more complicated, actions have reactions, and unpleasant consequences can follow. This, too, is plainly-enough cause and effect. 

Let’s return to Morgan Stanley. On Tuesday, after the market closed, the wire services carried the following, “Morgan Stanley downgrades XYZ.” On Wednesday, nobody can be found to buy XYZ at the previous day’s price. Is that an answer to the question “why”?

We can say, I think, that it is on its face a candidate for cause. When we speak of “Morgan Stanley” we refer to various individual humans who, by law and convention, make use of that brand. Those people took certain actions; they pressed certain keys on their desktop computers. Those actions had immediate consequences such that for example a press release appeared on the screens of other computers. Those consequences may well have lessened the interest of some potential buyers of XYZ.