‘Real’ evidence is hard to come by

I’ve just been complimented on Twitter for offering ‘real’ evidence.  While it’s pleasing to be flattered,  I ought to add two cautionary notes about my last blog.  The first is that I started the entry by expressing reservations about the direction of causation.  It’s a problem that bedevils most statistical analysis – showing an association between two factors often looks plausible, but it’s not possible to say confidently whether there is a direct relationship, if the connection is best explained  through another factor, or there’s some further aspect of the figures I’ve been using that produces an artefact.  (The figures I had to hand, as it happens, were mainly for larger countries.  I don’t know whether that matters.)

The second reservation is the curse of comparative data – the graph is based on a handful of observations, where each national government counts for one uniform policy unit (e.g. the USA, Switzerland, Denmark), regardless of differences in size, variations in structure and rules, and so forth.  Belgium has at least six governments and I don’t know how to count Switzerland.   The numbers, Castles argues in Comparative Public Policy, allow us at best to run “a preliminary sorting process, informing us about combinations of variables which fit together to produce possible accounts”.

I suspect that the main determinant of labour market participation is the structure of the economy,  and while I’m not sure it has anything to do with benefits, there are hidden elements of conditionality in most systems which the crude figures won’t reflect.  That’s also why I finished up by saying, ‘if there’s an association…”

Leave a Reply