Inputs, outputs and incentives

I was reading a discussion on a blog somewhere recently and I read someone make the apparently uncontroversial point that obviously we need to reward people for out outputs not inputs. This is a mini version of the idea that often too much emphasis goes into what we put into a system (be it money or whatever else) without considering whether it is making the delivery better. This is of course a regular argument of people on Right about public spending.

But just in the more scaled-down version of incentives, I think the argument put forward (focus on output) is wide of the mark. In fact it might be much better to tie rewards to inputs in some cases. I blogged previously about Roland Fryer's work on educational incentives, and this is exactly the conclusion he came to. If you try and reward kids for grades it doesn't work (he argues that this might be because they don't know how to get the results). But if you reward them for inputs - read a book and you'll get $5 or whatever - then you can improve performance.

The conclusions to Fryer's report are pretty clear:

Remarkably, incentives for output did not increase achievement... Conversely, incentives can be a cost-effective strategy to raise achievement among even the poorest minority students in the lowest performing schools if the incentives are given for certain inputs to the educational production function.

From all the stuff I have read over the past year or two on incentives and motivation, one message about their success comes across clearly. The task needs to be well-defined, measurable and pretty basic. That suggests to me that incentives simply aren't that useful for many jobs, where quantifying 'good' performance would be difficult (and this is without even going into motivation crowding theory). But if we really do want to use incentives, it might actually be worth seeing what inputs correlate with good performance and applying them there. It may still not work, but tying rewards to results is not actually the no-brainer people assume.