Friedman's Methodology: Maximally Questionable.

The expression below shall prove useful, even if we avoid the maths. So, be brave:

        Max p*q(K, L) – w*L – r*K
       K, L
(1)     K >= 0
        L >= 0

Firms and households/consumers are the two fundamental economic agents: microeconomics begins with their study. Expression (1) is the firm's most basic representation: it shows its profit maximisation problem in the long run and embodies a number of different assumptions (see Appendix 1, for details). Even if readers find that expression otherwise cryptic, at least the "Max" should indicate "maximisation", while p*q - w*L - r*K is profits: revenue (price times quantity) minus costs (wage bill: wages times manhours; and capital costs: percent return times capital).

A similar expression holds for households: in a way "firms" maximising a "profit" measured in utils (see Appendix 2). (Alternatively: firms are "households" maximising a "utility" measured in $, given some constraints: hint, try adding the Lagrange multiplier).

The assumptions mentioned have different rationales:

1. Some have an economic meaning (a price-taker, for instance, faces constant prices);
2. Some are included for mathematical convenience/tractability (generally, to get a unique real closed-form solution);
3. The maximisation assumptions are different and are particularly important for the purposes of this post: economic agents are supposed to behave that way (hence, we'll call these assumptions "behavioural"). They aren't derived from any other previous assumptions.

Some variables are measurable (at least in principle, if not in practice), some are not. All of them involve knowledge considerations.

----------

Milton Friedman's infamous 1953 paper "The Methodology of Positive Economics" touches several subjects, although the most important and controversial is his argument against realist assumptions.

Friedman discusses economic theories in abstract. That's not peculiar to him: it's common to methodologists of economics. It's, however, a very bad habit and my reasons to believe this should become evident soon enough.

Friedman distinguishes predictions from assumptions. The truth of a theory depends on the accuracy of its predictions: predictions, therefore, must be evaluated. Assumptions aren't to be evaluated; their realism is irrelevant. Their role is to provide a starting point from which the predictions are deduced.

A proposition, in other words, is either an assumption or a prediction. Sure, a prediction from argument 1 could be an assumption for argument 2, but within the same argument, it certainly cannot be both.

In abstract, that may not raise eyebrows: its absurdity, however, strikes any reader considering concrete economic models, like those I mentioned above. They embody several different assumptions, among them those we call "behavioural".

Take expression 1. Its behavioural assumption is that firms maximise profits. In other words, left to themselves in their "natural habitat", firms will maximise profits. Yes, that's right: that behavioural assumption is also a prediction! Likewise with expression 2 and the household's utility maximisation problem.

So, what to do with propositions that are both assumptions and predictions? Should they be evaluated or not? If they must be evaluated, how is the utility maximisation assumption/prediction to be evaluated? Wouldn't one need to measure utility? What if firms do not maximise profits, as assumed and predicted? Should Friedmanite economists be free to evaluate only those propositions they find convenient?

This deficiency becomes particularly embarrassing and damaging for Friedman's methodology if one considers that those two models are indeed fundamental for the whole of microeconomics.

----------

Two final notes: I don't claim originality for this observation. That it may surprise some discussants of Friedman's methodology is probably more indicative of their limitations than of my intellectual prowess (readers, however, are subtly encouraged to disagree).

In fact, the paper has a few obvious wrinkles and in the future I might treat some of them, touching also more recent controversies.



Appendix 1:
Two-input, single-output, profit-maximising firms. Assumptions

1. Firms are price-takers and know their prices: p (price in $/unit), w (wage in $/manday or manhour), r (capitalist's return %) are strictly positive real constants. This means that firms' activities cannot affect the prices it sells its output, or the prices it buys its inputs: they are exogenous.
2. All factors are mobile: K and L (capital in $, and labour in mandays/manhours) are real continuous variables (inputs are infinitely divisible).
3. The firm produces only one output q(K, L) which is measured in units (continuous, twice differentiable, monotonously increasing, second derivatives negative). These assumptions ensure the problem has a unique closed-form solution.
4. Firms maximise profits.

Appendix 2:
Two-good, utility-maximising households. Assumptions

        Max u(X, Y)
       X, Y
(2)     px * X + py * Y <= M
        X >= 0
        Y >= 0

1. Households are price-takers and know their prices and budgets: px, py (prices in $/unit of X or Y), M (budget in $) are strictly positive real constants. This means that households' activities cannot affect the prices it buys its consumption goods. Prices and budgets are exogenous.
2. Goods are infinitely divisible: X and Y (goods in units) are real continuous variables.
3. u(X, Y) is measured in utils (continuous, twice differentiable, monotonously increasing, second derivatives negative). These assumptions ensure the problem has a unique closed solution.
4. Households maximise utilities. They can order consumption bundles according to their preferences.

UPDATE:
19/11/2016. Some typos/repeated text corrected in Appendix 2.