It is a major bone of contention of mine that the word ‘assumption’ is used interchangeably when in many cases it should be replaced with ‘hypothesis’ in economics – for example, that firms equate MR=MC is a hypothesis that can be falsified in its own right, rather than an ‘assumption’ in the purely scientific sense of the word.
Economists enjoy demonstrating that they don’t understand the difference between a good and a bad assumption. For example, here are the SuperFreakonomics guys:
There are some 237 million Americans sixteen and older; all told, that’s 43 billion miles walked each year by people of driving age. If we assume that 1 out of every 140 of those miles are walked drunk — the same proportion of miles that are driven drunk — then 307 million miles are walked drunk each year.
Convenient if you can’t be bothered to do your research, but scientifically worthless. This is a hypothesis about how people behave, and the analysis follows directly from there. If the hypothesis is wrong, the analysis is simply wrong and we need to start over.
Now, here’s Scott Sumner on the Diamond and Saez ‘Marginal Tax rates’ paper:
And S-D also seem to lean toward the “assume a can opener” school of policy analysis:
“In the current tax system with many tax avoidance opportunities at the higher end, as discussed above, the elasticity e is likely to be higher for top earners than for middle incomes, possibly leading to decreasing marginal tax rates at the top (Gruber and Saez, 2002). However, the natural policy response should be to close tax avoidance opportunities, in which case the assumption of constant elasticities might be a reasonable benchmark.”
So there you are. It’s just too much to ask of our policymakers to actually make hedge fund managers pay labor taxes on their labor income, but S-D have no problem waving a magic wand and assuming away all tax loopholes.
Of course, this is perfectly good assumption from a scientific point of view, as the presence of tax loopholes has a fairly simple (albeit hard to calculate empirically) impact on a variable, e. We can easily adjust the analysis to change this later on.
I feel it is important that, to progress, we need to differentiate between assumptions, for which a relaxation has a clear mathematical impact on the analysis, and hypotheses, which themselves need to be empirically verified, and for which a relaxation causes a model to collapse completely.