This is just the result of a tired brain running on empty a thought experiment - it doesn't necessarily reflect my beliefs, but it might be something I'm worried about. I'm going to be loose with lingo here, so apologies in advance:
Imagine that there is a production function for improving the welfare of poor people, and it has five inputs, A B C D E. As policymakers, we know that there are diminishing returns to each input in turn, and we know that the inputs are complements (the marginal impact of each input increases as the other inputs increase), but we don't know the exact rate of return on each input. For some, it could be close to zero - and for some it could be massive.
In the state of no knowledge and with constrained resources, we hedge our bets and divide our resources among the five inputs, hoping we are making a difference to the welfare of poor people.
Suddenly, someone comes along and randomly reveals the marginal impacts of intervention A and B, and we learn they are massive and nil, respectively - but we don't know anything about C, D and E (you could say it's either too expensive or politically unfeasible to reveal this information).
The question for the evening is, as a policymaker, do you:
- Put all of your funding in A and withdraw all of it from B-E.
- Only withdraw it from B, put most of it in A but a little in C-E (perhaps you have an idea of what distribution the marginal product of C-E is drawn from, so you fund each accordingly to maximize expected welfare).
6 Comments
I would say it is a truly pathetic state of affairs that every person who needs one does not already have a bednet, that this is a massive open goal, that it is a massive collective failure on the part of humanity, and we can do better.
I think I've missed your point. Are you saying that you favor 1)? Or are you taking some sort of Gates-type stance that we should prioritize something, fund it to saturation, then move on?
Whoa. After my own experiences in poorly-governed country (or at least, low-capacity governance) this maybe answers why some ridiculous paths are followed.. As this theoretical policymaker, how do you tell your constituents (tax base) that you are now doing 2) - spending money on things that you can't reveal or can't quite measure yet? I can see media jumping on the failure of activity B, the success of A, and popular opinion wanting nothing but A. Eeeesh.
Hi Deg, nnHmmm - I'd say you're being incredibly optimistic about the public's ability to weigh these things. I'd say close to 95-99% of national budgets of most countries go towards things that haven't passed a rigorous impact evaluation (and that's the only thing that can reveal the true marginal impact of A). I also doubt the media has much of an idea of what `proven' really means - this (at least in the UK) is still a media which reports the results of seriously flawed medical studies revealing X is correlated with Y. These are even constituents which tolerate seriously-unproven things like homeopathy in UK and witch doctors in several SSA countries. nnYou justify spending on C D and E (remember option 2 also removes all funding for B), by the very fact that there is uncertainty there. Even if you are completely risk-neutral, it's still rational, and optimal, to keep funding these things as long as the expected return is positive, based on your best information.
Isn't this a variation on the Monty Hall problem? http://en.wikipedia.org/wiki/Monty_Hall_problem If so it's probably worth ramping up your investments in C to E too possibly randomly choosing one of them to spend most of your money on
In reality you are likely to be dealing with a complex situation in which the relationship between the different factors is non-linear and interdependent and probably also varies over time. I'd suggest finding some way to experiment with different proportions of funding to each and iterating it over time based on those combinations that work best - although I doubt that this approach would be politically feasible.
I think you've pretty much answered your own question in your second comment, Matt. Presumably, the reason why investment in B is nearly nil is because there is a binding constraint in A, C, D or E or all of them that needs to be addressed before B's MR increases again. We know A has a high marginal return, indicating (given that an increase in any of A-E increases MR of the others) that it is underinvested and will then increase MR of B-E.
However, the information we have doesn't tell us anything about the evolution of the MR curve of A given current levels of B-E. We might find high marginal returns that decline sharply based on constraints in C-E. Or we might not reach that point for a while.
Basically, my point is that static MR and impact analysis is still insufficient for appropriate budgeting, except in the negative. It's more useful for budgeting to know something has 0 MR than to know it has very high MR, because we need to know how that MR evolves given constraints elsewhere. We have to rely on our largely qualitative assessment of where constraints are, and our relatively closeness to the binding constraint in any input.
To varying extents this is how any Government deals with budgeting. Noise comes from the lack of good information / data; lack of clear budgeting policy; political intervention; and sometimes madly distorted incentives.
Fascinating thought experiment, in any case.