Friedman’s Methodology: A Stake through the Heart of Reason

[[bit.do/weafm]] In “The Trouble with Macro”, Romer writes that macro-economists casually dismiss facts, and the profession as a whole has gone backwards over the past few decades, losing precious and hard-won knowledge. He does not consider WHY this happened. What are the methodological flaws that create the possibility of moving backwards, losing knowledge, affirming theories known to be in conflict with facts. How is it that leading economists can confidently assert theories which border on lunacy, and receive Nobel Prizes instead of psychiatric treatment?

This is due to the famous AS-IF methodology of Friedman, which gave economists a license for lunacy.  Friedman came up with this defense of orthodoxy when numerous emprical investigation revealed clearly that firms did not maximize profits, did not know their marginal costs, typically used mark-up pricing, and did other things which did not square with neo-classical theories.  Friedman argued that the validity of the axioms of a theory was not relevant; all that mattered was the ability to derive true conclusions from them. This has been called the F-twist:  “Truly important and significant hypotheses will be found to have “assumptions” that are wildly inaccurate descriptive representations of reality, and, in general, the more significant the theory, the more unrealistic the assumptions.” Friedman’s fallacious argument satisfied a deep-seated need — since the daily bread and butter of economics is ‘wildly inaccurate descriptive representations’ — and hence it became wildly popular and widely accepted throughout the economics profession. Now one can, without any embarrassment, make bizarre assumptions worthy of the lunatic asylum, and this is indeed the daily occupation of leading economists.

Friedman’s argument has been universally condemned by logicians and philosophers as an instance of the logical fallacy of “Affirming the consequent” – the use of modus ponens in reverse. That is, Friedman says, in effect, that theory T implies observable consequence C. We observe C, and therefore we can affirm that T holds. This is obviously fallacious since many different theories, inconsistent with T, may also imply consequence C. Even more importantly, a false theory T will always imply consequences C* which are not observed — since the theory is false, it will have consequences which are false. Ignoring all of these problems,  Boland uses an instrumentalist interpretation to defend Friedman, just like all economics textbooks. He writes that even though critics universally condemn his logic, Friedman is right, and ALL the critics are wrong.

In my lecture on  AM2L07 (code for Advanced Micro II: Lecture 7) Methodological Mistakes: Prospect Theory and Psychology Protocols, I explain why Friedman is wrong and his critics are right by discussing this methodological debate within the concrete context of trying to understand search theory. Consider a hypothetical problem where a person is searching for the highest wage. He goes from one firm to next. At each point he is offered a job at a certain wage W. He can accept and quit searching, or reject the offer and go on searching. We want to find a theory which explains search behavior that we observe in lab experiments designed to emulate this situation.

In simple models, it is easy to show that optimal search sets a reservation wage W* and the laborer searches until he/she finds the first offer above this value. The Economist is committed to the assumption that humans are hyper-rational, and they maximize. ONLY theories satisfying these assumptions will be examined for validity. This means that there is NO QUESTION of looking at human behavior itself to see whether or not this hypothesis about behavior holds. Rather, the ONLY problem is to find the FUNCTION which is being maximized.  Economists start by using Expected Utility theory. A rather large number of empirical examples show that this theory does not match human behavior. Nonetheless, this continues to be the dominant theory of decision making under uncertainty and continues to be taught in textbooks, even though the theory is KNOWN to be wrong.

An improvement upon this is PROSPECT theory. By making ad-hoc modifications to probabilities, utilities, and FRAMING the problem in a suitable way, this theory can achieve a MUCH BETTER match to observed behavior than Utility theory. This theory preserves MAXIMIZATION – humans maximize something. However it abandons rationality — why should humans treat probabilities INCORRECTLY. Economists cannot stomach this observed failure of rationality and so AFFIRM theories solidly in conflict with observed facts about human behavior.

NEITHER of these approaches is scientific, since both dogmatically assert allegiance to the maximization principle regardless of observation. The articles by John Hey show how one can move beyond this to a genuinely scientific methodology. He explains how many researchers have investigated search behavior, but have only been concerned with whether or not it matched ASSUMED theories of behavior. INSTEAD he proposes to investigate how humans ACTUALLY behave, without imposing any assumptions about behavior in advance. He used psychological protocols, asking subject to think out loud about the process with which they arrive at the decision on whether to accept an offer or to go on to search for the next one. As can be expected, humans cannot make complex calculations that theory requires of them, and instead they use various heuristics and rules of thumb. These heuristic work fairly well, and get them reasonable close to what someone with full information and infinite computational capabilities could achieve. Nonetheless, the use of heuristics gives radically different results about what we could expect to see in markets where these behaviors, rather than the hypothesized AS-IF behavior is used. The full lecture is linked below

For lecture slides and reference materials, links to related articles, as well as the whole sequence of lectures, see the course website: Advanced Micro II (shortlink: bit.do/ee2018)

POSTSCRIPT: The process of lecturing, trying to explain to my students how their fellow students are being duped by economic textbook, always give me greater clarity. In this lecture, I examine three approaches to understanding human behavior in the process of searching for the best wage (or searching for the best price).

1: AXIOMATIC — represented by Expected Utility. Here we know in advance what human behavior is. We do not need to look at human behavior at all. If someone ELSE studies this behavior and finds that our theories do not match actual behavior, we say that the experiment must be wrong.

2: DESCRIPTIVE — represented by Prospect Theory. Unlike economists, experimentalists and behaviorists study actual behavior. When it fails to match Expected Utility, they came up with a new theory — prospect theory — which summarizes and encapsulates a description of how humans behave in decisions under uncertainty. Economists REJECT this picture because it shows how human behavior is IRRATIONAL – and this conflicts with their FUNDAMENTAL assumptions of rationality, which must be maintained regardless of any inconvenient facts or observations or introspection.

3: SCIENTIFIC: An accurate description permits us to proceed to the next stage, which is to try to understand the REASONS for this behavior. For example, we observe that most people are risk-averse. They prefer the certain outcome of $50 to a gamble with offers $0 and $100 with equal probabilty. Now we can ask why — this is with regards to unobservable, hidden motivations, about which we can never be certain. A good explanation for this is REGRET. Because of our psychological makeup, the flatness of the utility function in gains, a win of 100 does not feel vastly superior to a win of 50. But the real kicker is the feat that if I take a gamble and lose, I will feel so stupid. Avoiding the regret that might occur when I say I should have taken that certain $50 might be the explanation for risk aversion.

Actually, even “reverse Modus Ponens” is not a good description of Friedman’s methodology — there is an added F-Twist: If we can FIND some observations C such that theory T implies them, then we affirm theory T, and IGNORE any other implications of T which actually conflict with observations.

The METHODOLOGICAL point is the Friedman, like all nominalists and instrumentalists, GIVES up on the possibility of understanding human behavior. All he wants is a model which provides a SUPERFICIAL match to some observations. However, many many real life situations show that this is NOT ENOUGH — we need to have a deeper understanding, in order to be able to explain economic and social phenomena.

See also previous related post on Economists Confuse Greek Methodology with Science. A more recent related post is: “The Methodology of Modern Economics“.