Lesson 7 of 7
What would change its mind
Asked of any theory, it is the fairest question there is. This tradition gives two answers, and one makes the question very hard to answer at all.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- One answer inside the school is that the cycle is a qualitative sequence detectable in the data only by applying the theory. On that view a statistical result cannot count against it.
- The other answer is the practice of the school's own empirical wing, which estimates models, reports coefficients and treats the results as support.
- Those two answers are not compatible. A reader is entitled to ask which one a given writer is using, and to notice when it changes mid-argument.
This is the last lesson of the course and the hardest one to write, because the honest answer is not a single answer. The question is the standard one: what observation, if it were made, would count against this theory.
Two answers given inside the same school
The theory-first answer
- The cycle is a recurring qualitative sequence, not a statistical regularity.
- It can only be detected in history by applying theory to it.
- So a finding of no cycle in the data is not a refutation.
- Advanced by writers defending the theory against statistical criticism.
The measurement answer
- Estimate the relationships the theory implies.
- Report coefficients, samples and periods.
- Treat a stable long-run relationship as support.
- Practised by the school's own empirical papers.
The first answer protects the theory from data. The second exposes it, and its own results are thin.
The theory-first answer, stated fairly
When Tullock objected that statistical tests find a random walk rather than a cycle, Salerno's reply was not that the tests were wrong. It was that the objection misses the methodological context: on this account the business cycle is not a mechanistic or statistical regularity that appears in history by itself, but a recurring qualitative sequence of abstract phenomena that can be detected in the data only by applying theory.
That is a coherent position and it has a real cost, which Tullock names in his rejoinder: if the word means something detectable only through the theory, then the theory is not talking about the thing most people mean by a depression, and a reader is owed that warning.
It is also the position that makes this lesson's question hardest. If no statistical pattern can count against the account, then the honest answer to "what would change its mind" is: not this kind of evidence.
The measurement answer, stated fairly
The school's own empirical work does not behave that way. Mulligan's paper estimates a relationship between real consumable output and the term spread on US monthly data, reports it, and treats a stable long-run relationship as support for the theory. Whatever the merits of that test, it is a test, and a different result would have been awkward for the author.
The same is true of the free banking literature from the third lesson of this course. Arguments there are conducted with reserve ratios, dates and failure counts, and one participant changes his mind about a historical case in print.
So the school does produce evidence claims. What it does not have is a settled account of what the evidence would have to show for the theory to be in trouble, and no writer this course read sets out such a criterion.
What a reader can do with that
Three things, and they are the practical residue of this whole track.
Ask which answer a given writer is using, and watch for the switch. A paper that presents a coefficient as support and then treats a contrary coefficient as irrelevant to a theory detectable only through theory has used both.
Separate the parts of the account that could be measured from those that cannot. Whether credit expansion precedes crises is measurable, and the answer is mostly yes. Whether the structure of production lengthens and shortens is measurable in principle and largely unmeasured. Whether individual errors cluster for the reason the theory gives is close to unmeasurable with existing data.
And apply the same test to the mainstream. A framework in which fluctuations are optimal responses to technology shocks has its own version of this question, and the next course asks it of every account it presents.
Check yourself
What is the theory-first answer to the question in this lesson?
- That the theory has already been confirmed
- That the cycle is a qualitative sequence detectable only by applying theory, so statistical findings cannot refute it
- That data before 1945 are unreliable
Show the answer
Answer: B.
Salerno's comment on Tullock states this position directly, in answer to the objection that the data show a random walk rather than a cycle.What cost does that position carry, on the critic's side?
- It makes the theory harder to teach
- The word depression then means something other than what most people mean by it, which the reader has to be told
- It rules out all historical evidence
Show the answer
Answer: B.
That is Tullock's rejoinder: he says the specialised usage is legitimate but that it should be declared, since the ordinary meaning is mass unemployment.Which claim in the theory is measurable in principle but largely unmeasured?
- That credit expansion precedes crises
- That the structure of production lengthens in booms and shortens in busts
- That people act purposefully
Show the answer
Answer: B.
Testing it needs output or employment by stage of production across countries and cycles, and that work has not been done at the scale the question needs.Sources
- Salerno, 'Comment on Tullock's Why Austrians Are Wrong About Depressions', Review of Austrian Economics, volume three: the position that the business cycle is a recurring qualitative sequence detectable in historical data only through the application of theory.
- Tullock, 'Reply to Comment by Joseph T. Salerno', in the same volume: the rejoinder that a specialised meaning of depression should be declared, since the ordinary meaning is one of mass unemployment.
- Mulligan, 'An Empirical Examination of Austrian Business Cycle Theory', Quarterly Journal of Austrian Economics, volume 9, number 2, 2006: the estimation and the treatment of a stable long-run relationship as support.
- Mises, Human Action, 1949, part one, chapter one, section three, where the aprioristic and deductive character of the reasoning is set out and defended against the objection that deduction adds nothing to knowledge. That is the methodological position from which the theory-first answer follows.
What this leaves out
- This is a heterodox tradition. The mainstream account of the same phenomenon differs, and nothing here is a forecast or a view on any security.
- No writer read for this course states a criterion that would count against the theory. The absence is reported as an absence; it is not evidence that none exists in a work this course did not obtain.
- The same question is asked of every rival account in the next course. It is not a test this track applies to one side only.