John Quiggin wrote an interesting post about the failure of search theory's prediction that with the advent of the internet (reducing search costs and times), unemployment should drop faster. I commented on the post, citing my attempt at matching theory using the information transfer model. I added that because the model depends only on the number of hires and vacancies and does fairly well completely ignoring any microfoundations, search costs and the internet may have nothing to do with it.
The comment thread has many "just-so" stories (or random rants about the government, economists or the plutocracy), however one commenter had an interesting take from queueing theory: the delay is proportional to the buffer size. With the internet also increasing everyone's buffer size, the gains from reduced cost and search times are eaten up and it is a wash. This is actually a measurable thing (the number of candidates businesses go through could be surveyed), so maybe it could have some explanatory power.
My opinion is that searching for a mechanism is something of a Sisyphean task. The thing is: recoveries in employment have remarkable regularity. I've noted the regularity before; here is another graph where I've excised the unemployment increases and show lines of the same slope (fit to the current recovery):
This regularity over several decades would imply that any mechanism that explains the rate likely has nothing to do with the internet, inequality, jobless recoveries, war, government spending, unemployment benefits, Keynesianism, monetarism, technology, ... etc. It is doubtful these different forces conspire in differing degrees to achieve approximately the same result every time.
I also wanted to repost the "metastability" I observed in the unemployment rate I talked about in this post just for fun.