Lesson 7 of 8
Pattern instead of model
Some accounts of crises are built from episodes rather than equations. What that method can establish, what it cannot, and how to read a claim built that way.
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
- The method is comparative history: assemble many episodes, look for a recurring shape, and describe it, rather than deriving it from a model.
- Its strength is coverage. A dataset spanning centuries and regions catches patterns that a sample of post-war quarters cannot.
- Its weakness is the counterfactual. A recurring shape does not say which of several mechanisms produced it, and selection of episodes is a judgement call.
This lesson is about a way of working rather than a theory. It belongs in a course about rival accounts because the strongest empirical result in the whole track, which is the next lesson, came out of this method rather than out of a model.
Two ways of establishing a claim about crises
From a model
- Derive the mechanism from stated assumptions.
- Produce quantitative predictions.
- Test against a sample where the variables are measured.
- Fails when the assumptions are unmeasured or untestable.
From episodes
- Assemble many cases across countries and centuries.
- Describe the recurring sequence.
- Check how widely it holds and where it does not.
- Fails when the counterfactual is unavailable or the cases are chosen after the fact.
A model says why, conditional on its assumptions. A history says how often, conditional on its sample.
Where the tradition of this method sits
The best-known book-length statement of the historical approach to manias and crashes is Kindleberger's, cited in the BIS survey used elsewhere in this track in its 2000 edition. This course did not obtain it, so nothing is reported here about its argument, and it is named for one narrow reason: that survey names Kindleberger and Minsky as the writers outside the mainstream in whose accounts financial booms and busts continued to feature while the subject was out of fashion.
That is a fact about where the method was kept alive, and it can be stated without characterising a book nobody here read.
A worked example of the method
Reinhart and Rogoff's 2008 paper is the clearest case a reader can check. It assembles a dataset of financial crises spanning from a fourteenth-century English default to the United States subprime crisis, across all regions, including credit episodes rarely covered in the literature.
Their stated findings are of the kind this method produces. Serial default is close to universal among countries in the course of becoming advanced economies. Major default episodes are typically spaced years or decades apart, which creates an illusion among policymakers and investors that the current period is different in kind. Crises frequently emanate from the financial centres and transmit through interest rates and commodity prices.
Notice the form of each. They are claims about frequency and recurrence, not about mechanism. Nothing in them says what causes a default, and the authors present the work as an illustration of what a sweeping historical database makes visible.
How to read a claim built this way
Three questions, and they apply to any historical pattern claim including the ones made inside this track.
Which episodes were included and on what rule. A pattern found in cases selected because they fit is not a finding.
What is the denominator. A list of crises that followed credit booms says nothing until you know how many credit booms were not followed by crises.
And what would the same method say about a period nobody has looked at. A pattern that only ever gets applied backwards is a description of a sample rather than a claim about the world.
The next lesson is what happens when this method is applied with an explicit sample, an explicit denominator and a published dataset. It is a much stronger result, and it is stronger precisely because those three questions have answers.
Check yourself
What kind of claim does the comparative historical method produce?
- Claims about mechanism, derived from assumptions
- Claims about frequency and recurrence, conditional on the sample of episodes
- Claims about optimal policy
Show the answer
Answer: B.
Reinhart and Rogoff's findings are of this form: serial default is near universal, episodes are spaced apart, crises transmit from financial centres.What does the phrase about this time being different describe?
- A statistical test
- The belief, encouraged by long gaps between episodes, that the current period is different in kind
- A type of financial instrument
Show the answer
Answer: B.
Their paper presents it as a recurring illusion among policymakers and investors, produced by the spacing of major episodes years or decades apart.What is the denominator question?
- How many crises there were in total
- How often the supposed warning sign appeared without a crisis following
- How many countries are in the sample
Show the answer
Answer: B.
Without it, a list of crises that followed a warning sign cannot distinguish a reliable indicator from a common condition.Sources
- Reinhart and Rogoff, 'This Time is Different: A Panoramic View of Eight Centuries of Financial Crises', NBER Working Paper 13882, March 2008, abstract and introduction: the span of the dataset, the finding that serial default is nearly universal for countries in transition to advanced status, the spacing of major episodes and the illusion it creates, and the transmission of crises from financial centres through interest rates and commodity prices.
- Borio, 'The Financial Cycle and Macroeconomics: What Have We Learnt?', BIS Working Paper 395, December 2012, introduction, for the naming of Kindleberger and Minsky as the accounts outside the mainstream in which financial booms and busts continued to feature, and for the reference to the 2000 edition of Kindleberger's book.
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.
- Kindleberger's book was not obtained for this course. It is named only as the BIS survey names it, and no claim is made about its contents.
- The three questions in this lesson apply to historical claims made inside this track as well as outside it, and they are the reason the previous course could not produce a verdict.