Unformatted text preview: e profits3. So why do scientists make patently
false assumptions? It’s because these greatly facilitate the development of a tractable model. By necessity, we have to
assume away many aspects of reality because it’s impossible to build a useful model which captures every (or very
many) finest-grain details of reality.
Having made simplifying assumptions scientists (like ourselves ) build models as follows: Make
prediction(s) If predictions
“rejected” by evidence
then model is falsified If predictions “not
evidence then model
“validated” Assumptions relaxed
and once more subject
to testing For example, suppose we want to understand, explain and predict how much money individuals save (say) each month.
In reality, monthly savings depend on many factors such as: disposable income, family size, age, peer effects, expected
inflation/deflation etc. A model with every one of these determinants will be unwieldy, cumbersome and thwart our
efforts to understand saving. For traction, why not start by understanding, explaining and predicting the link between
savings and one of the determinants of savings (say, income)?
Assuming there’s a link between monthly savings and “income” we will generate testable predictions (after all this is
what separates science from poppycock and hogwash). Here are some testable predictions we could make (each of
these is a separate “model”): (a) monthly savings are proportional to monthly income; (b) monthly savings are inversely
proportional to monthly income; (c) monthly savings do not depend on monthly income; (d) monthly savings are
proportional to your average lifetime income; etc.
Each of these “income and savings” models can be tested against data4. For example, if we find that monthly savings are
not proportional to monthly savings, then we’d reject model (a) above; but if we fail to reject this prediction then model
(a) is partially “validated” (but not fully confirmed be...
View Full Document