So I finally decided to sit down and have a chat with the master, the incomparable father of the modern Austrian economics, Ludwig von Mises. By way of my daily Mises memo, which I have avoided for some five years, I started reading Human Action. Academically, it is like the most life-affirming music I have heard in a long time. Although there are many points that I would like to address in his work, I will refer to one in this post: the temporal relation between actions, featured in chapter 5 of Human Action.
The attempt has been made to attain the notion of a nonrational action by this reasoning: If a is preferred to b and b to c, logically a should be preferred to c. But if actually c is preferred to a, we are faced with a mode of acting to which we cannot ascribe consistency and rationality.[6] This reasoning disregards the fact that two acts of an individual can never be synchronous. If in one action a is preferred to b and in another action b to c, it is, however short the interval between the two actions may be, not permissible to construct a uniform scale of value in which a precedes b and b precedes c. Nor is it permissible to consider a later third action as coincident with the two previous actions. All that the example proves is that value judgments are not immutable and that therefore a scale of value, which is abstracted from various, necessarily nonsynchronous actions of an individual, may be self-contradictory.[7]
One must not confuse the logical concept of consistency (viz., absence of contradiction) and the praxeological concept of consistency (viz., constancy or clinging to the same principles). Logical consistency has its place only in thinking, constancy has its place only in acting.
To me, this is a very powerful series of statements. It is a powerful affront to the way that much introductory economics is presented, although I disagree with the Austrians that neoclassical economics doesn’t allow for it. In any case, this is exactly how economics should be taught, right from the get-go. How many times have we heard arguments about the transitivity of preferences?
Logical consistency has its place only in thinking, constancy has its place only in acting. This suggests an overlap. While logical consistency may overlap with constancy to some degree, it would never be 100%– as if it approached an asymptotic barrier that it could never breach. And in most situations, one would expect a measurable, even wide gulf between the two as realized in the world. I certainly think this type of thinking may challenge the utility of instrumental variables and econometrics, but perhaps not as far as Austrians think. The overlap between logical consistency and constancy ought to be enough to inform policy decisions, a concept, I know, that Austrians often reject.
This is ironic to me because Austrians themselves fall prey to this very trap of confusing logical consistency with constancy. More in Part II of my most heart-felt marriage proposal, perhaps the only one I may be privileged enough to make in this fleeting life.
First note that Mises is talking about transitivity and definitions of actions, which is a bit different than the modern approach of ranking consumption bundles (via preference relations). For our purposes I’ll define an action a as a consumer’s choice at time t. Mises says that only one action can occur at a single instant. Hence if we observe non-transitivity of actions, it could be because the underlying preferences have changed. So this kind of behavior is not self-contradictory, if we allow for changes in preferences over time.
Human Action was written in 1940, well before the modern classic The Theory of Value by Gerard Debreu was published in 1959. Debreu’s book axiomatized economic theory and pretty much all modern books follow the same structure of his book. See chapter 4, section 4 (page 5 of the ch4 PDF). Modern economics does indeed begin from the most basic assumptions about consumer preferences. They don’t mention this in intro classes, but it’s probably for the best. It’s for the same reason you don’t prove the existence of the real numbers in calculus 1. The students wouldn’t really understand it or appreciate it at this point in their education.
I don’t get the logical consistency vs. praxeological consistency thing. Is he just saying that his example about apparent non-transitivity of preferences creates a logical inconsistency but by noting the changes in preferences can explain the apparent non-transitive action, we observe action which is praxeologically consistent?
That is a bit ridiculous to me. Once you note that preferences can change over time, you redefine preferences as something that is defined at an instant of time, and then your logical system becomes consistent again. This is how mathematical models work. If we find a contradiction we redefine something to get rid of it.
If his point is just that we should make sure our model reflects reality, then yeah, he’s right.
BTW: Although I’m faulting Mises for some things, his work is obviously a precursor to the modern angle I’m criticizing from. Although I’m not sure how influential he was in the development of mathematical preference relation theory, It’s highly possible that his arguments changed it’s development to account for things like time dependent preferences.
I think you’re wrong that students wouldn’t appreciate that point in their education. It seems to me that such a simple rudimentary introduction (even for the symbols, Debreu’s chapter 4 isn’t exactly rocket science) would hardly be too much.
I’m not sure what you mean by the first paragraph — but I think I agree. I don’t think Mises is saying otherwise. If I understand your fourth paragraph correctly, I think you’re being too brusque. You’re not saying that Mises is wrong, you are just saying that your model can deal with it. Mises is saying that the assumption of indifference curves (and I am indifferent to the battle over this) may make logical sense, but very little in the real world.
Give it some thought. You don’t think so? I think his point is that models will never absolutely capture reality so that it is a faulty premise of modern economics that using such models will lead to the truth. I argue that whether or not it leads to absolute truth is irrelevant, as long as it clears up some distortion and/or conveys useful information. Surely we can bridge this theoretical gap.
I read through both Arrow and Debreu before my post actually, and had originally thought about discussing pareto-optimal outcomes, but am now having second thoughts.