I have long been a fan of Austrian Economics. In fact, possibly nothing has altered my thinking more than the work of the Austrian economists at the Mises Institute. However, I do not identify as an Austrian for two reasons:

  1. I do not know enough economics in general to assess if the Austrian school is the best
  2. I have several concerns with the Austrian methodology, one of which I will lay out in this blog post.

I first thought of this while listening to a debate between Bob Murphy and Vox Day on the resolution: free trade is always economically beneficial in the long-term and the more free trade is practiced by a country, the higher the standard of living of its inhabitants will be. While I absolutely support free trade, I thought the wording of this resolution to be particularly indefensible. Free trade is always economically beneficial in the long-term? For any given country and situation, free trade will lead to a higher standard of living for its inhabitants?

I can easily provide a hypothetical counterexample to disprove this resolution. Let’s say there are three countries: A, B, and C. Country A wants to open up trade with country B. However, country C — the only country with nuclear weapons — says that it will carpet bomb country A every year that it has free trade with country B.

In this scenario, the inhabitants of country A would be worse off by all the aforementioned metrics (assuming they are even still alive) if they establish free trade with country B — thus providing a possible, albeit unlikely, counter example to the resolution.

The Austrian Response

The Austrians have a way of handling this critique. When confronted with this possibility, they often bring up the concept of ceteris paribus — keeping all other conditions equal. By this logic, under which Murphy and Day were almost certainly operating, one can dismiss the above scenario by saying, given that the carpet bombing occurs, the inhabitants of country A are still better off being able to trade with country B than not.

And this is entirely reasonable — primarily because the resolution was solely focused on the effects and dynamics of free trade, and not on the effects of free trade and carpet bombing. In the case of assessing the benefits of free trade and a carpet bombing, an Austrian would likely say that the result is ambiguous because it is impossible to compare the utility gained from free trade and the disutility caused by the bombing.

Furthermore, Austrians are very careful in their wording. When, for instance, saying that the minimum wage causes unemployment, they don’t mean to say that the unemployment rate will actually go up. Instead, they recognize that there are many factors that influence employment and therefore claim that minimum wage will simply cause more unemployment than there otherwise would have been in the counterfactual scenario in which the minimum wage had not been enacted.

Response to the Austrians

The inspiration for this response comes from the excellent book, The Myth of the Rational Voter by Bryan Caplan (which I have, regrettably, not finished reading). In it, Caplan addresses a common argument in favor of democracy that goes something like this:

  • Let’s say 99% of people don’t know anything about the subject on which they are voting and 1% are experts
  • Therefore, probabilistically, 50% of those who don’t know about anything will vote yes, and the other 50% will vote no
  • This leaves 49.5% voting yes and 49.5% voting no, leaving the actual decision up to the 1% of experts
  • Thereby, leading to an expert-driven decision

Caplan, however, pretty convincingly destroys this narrative by demonstrating that the distributions are not 50/50 among the non-experts. He does this by showing that people systematically have biases when voting for policies and therefore cause sub-optimal outcomes. For instance (to Bob Murphy’s credit), voters have a natural affinity towards protectionism, even though most experts agree it is not beneficial.

So, how does this apply to Austrian Economics? Well, it is the Austrians that taught me the economy is complex. There are so many factors that it is impossible to model it with complete accuracy. The only way we can make any claim with any significant degree of certainty, is if we can induce it logically from the axioms of human action. However, I don’t see Austrians taking this to its logical conclusion.

Consider the case of the minimum wage. Again, I am opposed to the minimum wage, but am not completely on board with the Austrian rationale. The Austrian position, very briefly, is that, due to the price floor in the market, there will be unemployment among workers whose marginal revenue product is lower than the value of the floor.

While this I do not dispute the validity of this mechanism, this argument, as a whole, is what we in computer science refer to as a partial function.

The Logical Function

To illustrate why this is important, imagine that there is a set P of the state of all factors that possibly affect unemployment. As the Austrians point out, this set of factors would be so large, it would be impossible to accurately model mathematically. Now imagine that there is a function M, that takes two inputs — a given unemployment rate and a corresponding value of P — and predicts the effect of unemployment one year later, given all of this data.

Assuming U is the unemployment rate,

U1 = M(U, P)

P contains many variables, some of which are R, the rate of economic growth, S, the mindset of a small town’s monopsony-holding employer, and MRP, the marginal revenue product of the employees. Thus, the above statement can be expanded as,

U1 = M(U, R, S, MRP…)

For the sake of this example, let’s also assume the starting unemployment rate is 5%.

U1 = M(.05, R, S, MRP…)

All other variables in P equal, if MRP includes a significant amount of wages below the price floor, a plausible output might be,

.06 = M(.05, R, S, MRP…)

Conversely, all other variables in P equal, even with the prior MRP, if R, the rate of economic growth, is high enough, a plausible output might be,

.04 = M(.05, R, S, MRP…)

This shows that minimum wage doesn’t always lead to higher unemployment. If economic growth is high enough, unemployment can still drop.

But the Austrians have a brilliant response to this. They point out that R is unrelated to the minimum wage and can show that the minimum wage makes us relatively worse off compared to how we would have otherwise been without a high R. (For convenience, in the following example MRP as true indicates that there is an effective price floor, and false indicates that there is not)

Minimum wage + High R

A = M(.05, Rhigh, S, true…)

Minimum wage + Not high R

B = M(.05, Rnothigh, S, true…)

No minimum wage + High R

C = M(.05, Rhigh, S, false..)

No minimum wage + Not high R

D = M(.05, Rnothigh, S, false..)

In which case,

A > C and B > D

In other words, we would have a higher rate of unemployment with minimum wage and a high rate of economic growth than we would with that same high rate of growth and no minimum wage.

But this doesn’t always work. Imagine another scenario, wherein there is a monopsonist employer in a small town. This business owner hates automation and loves the minimum wage. Not only that, but he has very deep pockets and proclaims that if minimum wage is enacted in his town, he will hire more people and pay them even more. However, if it is not enacted, he will find a way to reduce employment by reducing his operation in the town and lobbying the local city council, with which he has extreme influence, to keep out competitors.

Given this value of S, the prior value of MRP, and all else equal, a plausible output might be,

.03 = M(.05, R, S, MRP…)

And this time, since the reduction in the unemployment rate is highly correlated to — and in fact, directly caused by — the implementation of the minimum wage, the Austrians’ prior technique does not work,

Minimum wage + Mindset

A = M(.05, R, true, true…)

Minimum wage + No mindset

B = M(.05, R, false, true…)

No minimum wage + Mindset

C = M(.05, R, true, false…)

No minimum wage + No mindset

D = M(.05, R, false, false…)

In this case, it is not true that

A > C and B > D


A < C and B > D

In other words, if this mindset is in place, unemployment will be higher without the minimum wage than it would, under the same circumstances, with the minimum wage.

However, the Austrians still have a pretty good response. Think back to the earlier example of the countries going to war over free trade[1]. It’s fair to assume that such an obscure and improbable situation is irrelevant to the discussion. If presented with this monopsony situation, an Austrian would probably say it’s indeterminate, since it’s possible that the business owner could die or another firm could uproot his stronghold in the town.

Further, and more importantly, this is a one off situation and, thus, generally holding the position that the minimum wage causes unemployment is still usually correct — and therefore a good conclusion. After all, B > D and situations A and C are very rare.

The X-Factor

Now imagine that, in our analysis of the labor market, when we think about the way that employers respond to price and productivity, there is a factor we have not considered. Although this seems like a mistake we would not make — many people do make it. A lot of people, for instance, look at the labor market and think only of the supply side, assuming that the cost of employment is not a factor in the minimum wage’s effect. The Austrians, seeing the bigger picture, point out that, if the cost of hiring a worker is higher than his or her productivity, the worker will likely be fired.

But what if the Austrians themselves are forgetting to include a factor in their logic function. In other words, what if their logic function is incomplete.

In the case of the small town, we showed a situation wherein, ceteris paribus, minimum wage would cause less unemployment than no minimum wage. But does that small breach in logical completeness invalidate a whole theory that otherwise is completely sound? I’d say yes.

Imagine that there is a factor that we will call X. This factor is strongly correlated to, and is, in fact, caused by, the minimum wage. However, X is not taken into account the Austrians’ analysis. X would have a similar breakdown to S in the previous example and therefore,


U1 = M(U, R, S, MRP, X…)

Minimum wage + X

A = M(.05, R, S, true, true…)

Minimum wage + No X

B = M(.05, R, S, true, false…)

No minimum wage + X

C = M(.05, R, S, false, true…)

No minimum wage + No X

D = M(.05, R, S, false, false…)

In this case, it is not true that

A > C and B > D


A < C and B > D

Finally, imagine that X occurs as often as S doesn’t occur. Or, alternatively, it is a valid consideration as often as the standard Austrian position comparing productivity to cost — neither of which applies 100% of the time. In this case, arguing that, ceteris paribus, B > D, and we shouldn’t consider A or C, doesn’t make sense. Just like Caplan showed that, although logically valid, that common conception of democracy, in practice, does not work due to systematic biases; factor X, in practice, systematically alters the effect predicted by the Austrians.

Maybe factor X doesn’t exist. Or maybe there are factors X, Y, and Z that come into play. The fact is, we can never know. As far as I can tell, there is no way to prove the completeness of the Austrian position, so long as P is too large to mathematically model.

But this doesn’t prove that Austrians are wrong. In fact, I believe there is something very fundamentally right about their approach — a humility unparalleled to anything I have seen thus far. But that said, by their own logic, it is impossible for anyone to draw useful conclusions about the world from their approach. There can always be an x-factor and we can never (formally) check if the theories have significance in real life, given the Austrian prohibition on statistics.

Riddle Me This

Given the Austrian axioms, all else equal, one can “prove” that the minimum wage causes unemployment. However, how do you know that these axioms are a sufficient foundation to predict outcomes, even relative to their counterfactual scenarios, in the real world? Is Austrian Economics even useful at all?

Clearing up a few potential misconceptions:

  • I know the Austrian School doesn’t have a normative position on the minimum wage, my point is that one cannot reliably determine a position based solely on praxeology.
  • I am not deluded enough to think that I have come up with a unique criticism. I’m sure this has been brought up and debated before. However, I am phrasing it in my own way and, hopefully, wording it in a manner that contributes to the discussion. I would be genuinely happy to receive a devastating refutation (even if written originally for someone else, long ago). But, also, I do think I am correct.
  • I am not endorsing other forms of economic analysis either, although I am growing partial to the Chicago School. I have a lot to learn in economics and, for now, advocate solely for extreme humility.
  • This criticism, as far as I can tell, applies to all conclusions made by Austrians, including free trade and the business cycle. It is not solely targeted toward the minimum wage.
  • The free trade / carpet bomb scenario is not a proper application of the x-factor for free trade. It is a counterexample to the resolution debated by Murphy and Day.

[1] The free trade situation, as described in this post, and the minimum wage “mindset” scenario are slightly disanalogous because, in the minimum wage scenario, the contingent factor is directly changing the metric in question. Thus, a carpet bombed country is still benefiting from free trade relative the same carpet bombed country without free trade. However, since the monopsonist fully controls the outcome of the unemployment rate, it is not possible to do the same comparison without stipulating that the monopsonist dies, changes his mind, or loses control. One could say, given the monsonist’s decision to hire more people or not, with a minimum wage or not, the lack of the minimum wage makes the situation even better, however, this would inaccurately represent the reality of the situation — since it would be disassociating the effect from the cause.