A paper from the School of Continuous Improvement works out what a significance test can and cannot establish about a law stated as a tendency with a list of exceptions attached.
A paper on the falling rate of profit finds a regression can confirm the law and cannot refute it
We began by asking how large a sample we would need, and the answer for one of our cases was three hundred and five years. I am grateful to the School for the time it has given us so far.
— Mr Eskil Nwabueze, Doctoral Candidate, School of Continuous Improvement
The sample in question belongs to Testing a Tendency: Identification and Power for a Law That Lists Its Own Exceptions, published today by Mr Nwabueze and Dr Torun Ezeigwe, Senior Lecturer and Director of the Master of Applied Measurement. The paper takes the Marxian law of the falling rate of profit as it is classically stated, as a tendency accompanied by six counteracting influences, and asks what a statistician would need in hand before saying that the law held, or that it did not.
Its method brings identification analysis, measurement reliability and study planning to bear on a single regression. The authors find that the regression can only understate the tendency, so that a result in the law’s favour stands whatever the quality of its controls, while a result against it depends on a quantity no record of any length supplies. In a simulated economy where the law is true by construction and nothing is controlled for, a century and a half of annual figures reported the opposite in 99.7 per cent of samples.
Dr Ezeigwe said the School had “always held that a question deserves to be costed before it is answered, and it is pleasing to see one costed this thoroughly”.
The School counts the paper among the returns on its long habit of asking what a method is entitled to conclude, and is glad to see that habit carried into a literature the School had not previously had occasion to visit. The paper reports a mathematical analysis and simulated data only.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.5fu1p5.