Lesson 7 of 7
The decisive evidence
What would favour one account over another, whether it can exist, and why the economists themselves did not produce a winner. This course does not declare one.
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
- Each account implies something the others do not, and those implications are where a test would have to bite.
- Most of the tests are obstructed, not by laziness but by the data: the production structure is unobservable, the counterfactual monetary policy was never run, and expectations were never surveyed.
- No account in this course is declared correct here. The economists who spent careers on it did not converge, and a curriculum that resolves it in four pages is not teaching.
This is the lesson the whole course exists for. Not which account is right, but what would have to be true for each to be favoured, and whether anybody could find out.
What each account implies that the others do not
The distinguishing implication
- Austrian: resources shifted towards stages far from consumption during the 1920s and reversed after.
- Monetarist: output follows the money stock, with money leading rather than responding.
- Keynesian: expected yields on new capital stayed too low for any feasible rate to offset.
- Bernanke: credit costs rose for borrowers without securities market access, beyond what the money stock explains.
Why it is hard to test
- The stages of production are not observable, so the test rests on a chosen industry classification.
- Money and output move together, so leading and responding are hard to separate in a single episode.
- Expectations of yields in 1931 were never measured and cannot be recovered.
- Separating a credit-cost channel from a money channel requires series that mostly move together.
Every one of these tests is obstructed by something about the data rather than by a lack of interest.
What each account would need
The Austrian account needs evidence of the reallocation. Not that credit expanded, which the other accounts can accept, but that investment moved towards the stages furthest from consumption and reversed. The previous courses established why that is hard: the structure is a heuristic, not a series, and any proxy is a judgement made by the person testing.
The monetarist account needs the direction of causation. Friedman and Schwartz address it by identifying monetary changes not attributable to contemporaneous cyclical conditions, and by pointing to the timing of turns in the rate of change of money. Both are inferences from a record in which the two series move together.
The Keynesian account needs the expectations. The claim is about what investors expected new capital to yield, and nobody recorded it.
Bernanke's account is the one that offers the most direct test, which is part of why it has done well. He assembles a monthly series for the deposits of failing banks, the liabilities of failing businesses, bank lending relative to personal income, and the Baa to government bond spread, and asks whether these add explanatory power over monetary variables alone.
The evidence that cannot exist
Counterfactuals. The claim that the Federal Reserve could have prevented the collapse in the money stock, and almost surely without threatening the gold standard, is a claim about a world that was not run. It can be argued from institutional detail and from the size of the Federal Reserve's own holdings, and it cannot be observed.
The same applies to the Austrian claim that a faster correction would have shortened the depression, and to the Keynesian claim that a sufficient volume of directed investment would have restored employment. Each names a road not taken.
Why one episode cannot settle it
Because every account was built with this episode in view. An explanation constructed to fit a case and then tested on that case is not being tested.
The honest route is out of the case: many episodes, many countries, and the international evidence that the later literature developed, none of which was read for this course and none of which is described here.
What a reader should take away
That the argument is live, that the four accounts are not symmetric in how much of themselves they expose to evidence, and that the one which states its own limits most clearly is also the one that has aged best. Those three facts are worth more than a verdict.
Check yourself
What does the Austrian account need to establish that the others need not?
- That the money stock fell
- That resources shifted towards stages far from consumption during the 1920s and reversed afterwards
- That banks failed in waves
Show the answer
Answer: B.
Credit expansion on its own is compatible with the other accounts. The reallocation is the distinguishing claim, and the structure is not observable.Why can the claim about what the Federal Reserve could have done not be observed?
- Because the records were destroyed
- Because it is a claim about a counterfactual, a world that was not run
- Because the Federal Reserve was not yet established
Show the answer
Answer: B.
It can be argued from institutional detail and from the size of the holdings involved, and the Austrian and Keynesian accounts each contain a counterfactual of their own.Why can this single episode not settle the argument?
- Because the data start too late
- Because every account was built with this episode in view, so fitting it is not a test
- Because the accounts agree about everything
Show the answer
Answer: B.
The honest route is out of the case, into many episodes and many countries, and that literature is not described in this course.Sources
- Bernanke, 'Non-Monetary Effects of the Financial Crisis in the Propagation of the Great Depression', National Bureau of Economic Research working paper 1054, January 1983, for the monthly series he assembles in table 1 and for the question of whether credit variables add explanatory power over monetary ones.
- Friedman and Schwartz, 'Money and Business Cycles', Review of Economics and Statistics, 1963, reprinted in The State of Monetary Economics, National Bureau of Economic Research, 1965, for their treatment of causation through monetary changes not attributable to contemporaneous cyclical conditions, for their timing evidence, and for the claim about what the Federal Reserve could have done.
- Keynes, The General Theory of Employment, Interest and Money, 1936, book VI, chapter 22, for the claim about expected yields that a test would have to reach.
- Hayek, The Pure Theory of Capital, 1941, preface, and Lachmann, Capital and Its Structure, 1956, chapter 1, for why the production structure resists measurement.
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.
- The study described in the example does not exist and is sketched to show what would bear on the question. No result is claimed for it.
- This lesson does not declare a winner and no lesson in this course does. The large later literature on this episode, including the cross-country evidence and the work on the gold standard, was not read for this course and is not described in it.
- Four accounts is a selection made for teaching. Others exist and are not treated here.