It's definitely, absolutely, a measurement error - just, well, how large a one is it?

As Paul Krugman pointed out, productivity isn't everything but in the long run it's pretty much everything. That productivity of labour is the largest, by far, determinant of future living standards. Thus the recent slow down in the labour productivity numbers is something to worry about. Most people getting this wrong, including Martin Wolf:

One possible explanation is mismeasurement. It is, and always has been, difficult to measure the impact of new technologies, particularly now when many services are free and many are provided, invisibly, from outside the US. Yet it is hard to accept that measurement suddenly became more difficult in 2005, when the US productivity slowdown began.

He mentions that only to dismiss it as not being the true reason. Yet it is the true one, Hal Varian is right, GDP doesn't deal well with free.

The various possibilities are that we're not having a technological revolution, we're just about to have one, the rich are taking all the gains, or we're measuring it wrong.

So, the example we've oft used. WhatsApp provides telecoms services to some 1 billion people. It takes the labour of some 200 people to do so. There is no price associated with this service. No advertising, no fee, therefore where it appears in GDP is a little odd. For the only thing we do see are the wages of those who provide it. We just don't see any consumption nor production value, only those costs.

The effect of this is that WhatsApp appears in our global economic statistics as a reduction in labour productivity. We've got labour costs, no associated production nor consumption, that's a decrease in productivity.

So, we've a system whereby 1 billion people getting free telecoms off the labour of 200 people is recorded as making us poorer, lowering labour productivity? That's madness, isn't it. It's also an obvious measurement error. Thus the answer to our productivity problem is measurement error, isn't it?

The only thing left to argue about is whether measurement is some, most or all of the problem. We'd say more than all of it ourselves but we're willing to listen to counterarguments.