Causal inference and policy

This is Joseph

I was reading this blog and I noted this paragraph:

But Brexiters have created a hermetically sealed logic. Every warning is dismissed as Project Fear, with the jeer ‘you can’t prove Brexit will make that happen’; every time a warning comes true, it is dismissed as Project Fear Mark 2, with the jeer ‘you can’t prove it was Brexit made that happen’.

One problem in causation is that few things happen for one reason.  A company chooses to locate (or re-locate) for a host of reasons.  That doesn't mean that one cause may not be important, or even sufficient.  It just means that complex policy outcomes are hard to prove on a case by case basis.

So instead you look at frequency and patterns.  If homelessness is increasing as policy changes occur then maybe that is a bad thing?  If a firm relocates business to Ireland, could it be that it was planning to do so anyway?  Of course.  But financial uncertainty might make the vehicle more popular and so the timing may well be linked to a policy change.

Of course, all of these associations may be confounded and many (if not all of them) will have other causes that contribute.  Look at dating -- how rare is it that one attribute in a partner is the only consideration in a relationship.  But that doesn't mean that a particular trait (say intelligence or charm) isn't doing a lot of heavy lifting.

It is a tough area for inference but we should look at rates and tendencies.  If the death rate is going up among younger age groups it doesn't mean opioids or bad traffic regulations are to blame for all of this change, but it is worth understanding there is likely a probabilistic factor underlying the policy change and outcome.