Lesson 8 of 8
What the tests actually found
Every published test this course could obtain, with its sample, its period, its method and its result, and nothing added at the end of the table.
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 strongest published regularity in this area is not about the structure of production. It is that credit growth predicts financial crises, and it is mainstream work.
- Tests built explicitly on this theory exist, are few, and are mostly run by writers who hold the theory. That is worth knowing when reading their conclusions.
- No test isolates the theory's distinctive claim from its rivals, because every rival predicts trouble after a credit boom too.
This lesson is a table and a set of cautions. It reports what each study sampled, over what period, by what method, and what it found. No verdict is added, because a verdict is exactly what the evidence does not yet support.
What each study looked at
- Countries in the widest sample · Drehmann and Yetman, quarterly 1970 to 201742
- Countries in the long historical samples · Schularick and Taylor, and Jorda and co-authors, 1870 to 200814
- Countries in the financial cycle sample · Borio, industrialised economies since the 1960s7
- Countries in the studies built on this theory · Mulligan, and the plucking tests: United States only1
The table
| Study | Sample and period | Method | Finding |
|---|---|---|---|
| Mulligan, 2006 | US, monthly, 1959 to 2003 | Cointegration and a vector error-correction model of real consumable output against the cumulative term spread | A stable long-run relationship, which the author reads as support for the theory |
| Dupraz, Nakamura and Steinsson, 2021 | US unemployment, 1948 to 2020 | Regressions of the amplitude of each phase on the phase before it | Contractions predict the following expansion; expansions do not predict the next contraction |
| Schularick and Taylor, 2011 | 14 countries, 1870 to 2008 | Long-run dataset of money, credit and output; crisis prediction | Credit growth is a powerful predictor of financial crises |
| Jorda, Schularick and Taylor, 2012 | 14 countries, over 200 recessions, 1870 to 2008 | Local projections conditioned on macro controls | More credit-intensive expansions are followed by deeper recessions and slower recoveries |
| Borio, 2012 | 7 industrialised countries since the 1960s | Frequency filters and turning-point dating of credit and property prices | A financial cycle averaging about 16 years, with peaks close to financial crises |
| Drehmann and Yetman, 2018 | 42 economies, quarterly, 1970 to 2017 | Horse race between credit gap measures as early warning indicators | No alternative gap outperformed the baseline credit-to-GDP gap |
Four cautions about reading it
The first is about what the rows measure. Only the first two are tests of anything specific to this theory. The other four are mainstream work on credit and crises, and they are in the table because they are the best evidence anywhere that credit conditions matter, not because their authors were testing this account.
The second is about who ran the tests. The studies built explicitly on the theory are almost all by economists who hold it, published in the journals that belong to the tradition. That does not make them wrong. It does mean there is no independent replication literature of the kind that exists for the factor research in value investing.
The third is that no test here separates this theory from its rivals. A credit boom followed by a crisis is consistent with the Austrian account, with Minsky's, with the monetarist reading and with Borio's financial cycle. The distinctive claim is about the structure of production, and measuring that requires sectoral data that almost nobody has used.
The fourth is about what the theory's own defenders claim for it. Mulligan's paper ends by calling the explanatory successes of the theory robust and its policy prescriptions unmistakable, in a paper whose test is a two-variable cointegration result. The gap between what was estimated and what is claimed for it is a thing to notice, in either direction.
Named, and not obtained
Four further studies are cited in this literature and this course could not obtain them: Keeler in 2001 and Carilli and Dempster in 2001, both in the Review of Austrian Economics; Bismans and Mougeot in the same journal in 2009; and Lester and Wolff, again in that journal, in 2013. Mulligan's survey reports Keeler as using standardised quarterly data for eight US business cycles and finding that monetary shocks caused cycles propagated through relative price changes, and reports Carilli and Dempster as arguing that correctly anticipating inflation does not remove the incentive to take a rate below time preference.
Those are summaries of summaries. They are listed so a reader can go and read the originals, and nothing in this course rests on them.
Check yourself
What is the strongest published regularity in this area?
- That the structure of production lengthens in booms
- That credit growth predicts financial crises across countries and over long periods
- That interest rates predict recessions two years ahead
Show the answer
Answer: B.
It comes from long-run cross-country datasets and has been replicated. It is also consistent with several competing accounts, which is the point of the third caution.Why is the credit and crises evidence not a vindication of this theory?
- Because the data stop in 2008
- Because several rival accounts predict the same relationship, and the distinctive claim is about the structure of production
- Because the studies were done at central banks
Show the answer
Answer: B.
Minsky's account, the monetarist reading and the financial cycle literature all imply trouble after credit booms. The sectoral claim is what would separate them.What should a reader note about the tests built on the theory?
- That they use the wrong statistical methods
- That they are few, mostly single-country, and mostly run by economists who hold the theory
- That they have all been retracted
Show the answer
Answer: B.
None of that makes them wrong. It means the independent replication that exists in other empirical literatures does not exist here.Sources
- Mulligan, 'An Empirical Examination of Austrian Business Cycle Theory', Quarterly Journal of Austrian Economics, volume 9, number 2, 2006: the data section, the cointegration and error-correction estimates, the concluding claims, and the summaries of Keeler and of Carilli and Dempster reported in its literature review.
- Dupraz, Nakamura and Steinsson, 'A Plucking Model of Business Cycles', NBER Working Paper 26351, revised June 2021, section two.
- Schularick and Taylor, 'Credit Booms Gone Bust: Monetary Policy, Leverage Cycles and Financial Crises, 1870 to 2008', NBER Working Paper 15512, revised July 2011, abstract and introduction.
- Jorda, Schularick and Taylor, 'When Credit Bites Back: Leverage, Business Cycles, and Crises', Federal Reserve Bank of San Francisco Working Paper 2011-27, October 2012, abstract.
- Borio, 'The Financial Cycle and Macroeconomics: What Have We Learnt?', BIS Working Paper 395, December 2012, section one.
- Drehmann and Yetman, 'Why You Should Use the Hodrick-Prescott Filter, at Least to Generate Credit Gaps', BIS Working Paper 744, September 2018, abstract.
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.
- This is not a survey of the literature. It is the set of studies this course could obtain and read in full, and four further papers are named in the lesson as unobtained.
- Every row reports what the authors found, in their own terms. None of the rows is endorsed here, and the absence of a verdict at the end of the table is deliberate.