Setting budgets

I went to a big science talk this week. The details aren't particularly important, but the discussion of the budget process was. The speaker contrasted an earlier regime, where he had had a budget that was deliberately set lower than the expected cost of his projects, with an over-run expected, with the current system he had to operate under, whereby budgets were set higher, but he was not permitted to exceed the stated sum.

It made me wonder which resulted in the lowest expenditure for a given project, and what the risk implications of budgeting were. Clearly if you say 'don't spend more than $10M', then there is a huge temptation to spend $9.95M regardless of what it actually costs to get the job done. And if you say 'I think it will cost you more than $8M to get this done - see how much more' then some project managers will take perverse pride in trying to prove you wrong and do it for $7.9M. So I am rather inclined to recommend the deliberately low model of bidgeting. However, it is important in this setting that there isn't an incentive for cutting corners that should not be cut. Sometimes it really does cost $8.5M to do it right.