Krugman did a fascinating post on this question yesterday:

I went back to something that was a hot topic not long ago, and will be again if and when the crisis ends: the apparent lag of European productivity since 1995. One recent, seemingly authoritative study is van Ark et al; and I noticed something that gave me pause.

In their paper, van Ark etc. identify the service sector as the main source of America’s pullaway — which is the standard argument. Within services, roughly half they attribute to distribution — roughly speaking, the Wal-Mart effect. OK.

But the other half is a surge in US productivity in financial and business services, not matched in Europe. And all I can say is, whoa! ...

First of all, how do we even measure output of financial services? If I read this BEA paper correctly, we more or less use “checks cashed” — or, more broadly, the number of transactions undertaken. ...

Volume of transactions seems like a massively problematic way to measure productivity, to say the least, since the big concern is that financial-sector growth has been fueled by companies' increasing sophistication at taking a cut from rearranging assets in ways that are of little social value. (And sometimes much, much less--I'm looking at you, CDOs.) Worse, often times there's an incentive to engage in multiple transactions when one (or fewer, or none) will suffice, or to engage in circular transactions, just to collect more fees. (See this funny-if-it-weren't-true op-ed from Michael Kinsley a few years back about Avis car rental.) Which makes measuring productivity gains this way especially perverse, since the very thing that often signals a low contribution to social welfare (i.e., a high volume of transactions) would drive the measured gains up.

--Noam Scheiber