Why are some recoveries short and others long?

Because of real factors, in a nutshell:

Using the recession recovery point equal to the month when private payrolls first exceeded their previous peak level, this paper argues that it was the negative secular trend in manufacturing jobs that was the most important determinant of the length and depth of the last three recessions/recoveries. This negative secular trend changed the layoff/recall pattern of jobs in manufacturing into permanent displacements, a malady that lengthened the recovery periods and that is not the explicit target of either traditional monetary policy or traditional fiscal policy. Using the ideas gathered from an examination of the US two-digit sectoral data for the US overall, attention turns to the recession/recoveries of the 50 US states in the last three national recession periods. Regressions that explain the lengths and depths of the recessions in 50 US states reveal the importance of construction jobs, but the most important predictor was manufacturing jobs: the greater the share of manufacturing jobs prior to the recession, the worse was the recession/recovery.

That is a new NBER working paper by Ed Leamer.  This of course bears on current monetary policy debates.  A very firmly held view on Twitter is that our post-2012 (or so) recovery could have been quicker, had the Fed been more aggressive, and thus we cannot afford to make the same mistake again.  Yet after a certain point it was real factors responsible for the recovery problems, and this new Leamer paper shows that (money still should be have been looser earlier on, in my view).  See also my previous coverage of papers by Marianna Kudylak (and co-authors), and Bob Hall directs my attention to this recent paper on why employment right now is recovering as fast as it is.  The evidence really is piling up very rapidly and decisively that mainly real factors are/were the problem after a certain point.