iS8 SECTIONAL ADDRESSES 



leading to a revolution of the kind which occurs from time to time in 

 physics. Or, more moderately, they may lay some stress on observation, 

 but urge that this should be done very much in the light of existing theory, 

 to test hypotheses directly suggested by that theory. 



Both schools must be given our cordial blessing. Past achievements 

 are still too exiguous for us to be sure which is the method most naturally 

 adapted to our study. 



It is sometimes claimed that the major part of established generalisa- 

 tions have been reached in the less empirical way. But my feeling is that 

 the great fruitfulness of the analytical map in yielding valid prescriptions 

 has obscured the extreme paucity of our knowledge with regard to causal 

 sequences. Two circumstances militate against the more deductive 

 method. One is the impossibility of the crucial experiment. In the 

 mature sciences which rely mainly on this method, such as physics, or, 

 to name a more recent comer, genetics, the crucial experiment is of central 

 importance. Secondly, it is extremely difficult to test hypotheses by 

 the collected data of observation. The operation of the plurality of 

 causes is too widely pervasive. Thus numerous hypotheses are framed 

 and never submitted to decisive test, so that each man retains his own 

 opinion still. 



I do not wish to press these considerations hard, but only sufficiently 

 to upset the complacency of dogmatic upholders of one exclusive method. 

 To give a contrary example, I believe that in so far as the monetary 

 explanation and the demand-for-capital-goods explanation of the trade 

 cycle be regarded as rival hypotheses suggested by theoretical considera- 

 tions, the course of events in this country and the United States in the 

 last ten years enhances the probability of the latter. It should be possible 

 to devise statistical methods to increase the cogency of this indication of 

 experience. I assume that even the more deductive or hypothetical 

 method of advance should be fortified by statistical verification. 



It is a doubtful point whether the more radically empirical method has 

 been as barren as is sometimes suggested. To give a rather trivial ex- 

 ample, Gresham's Law is an instance of the facts speaking. However 

 convincing the ex-post theoretical explanation of the phenomena, the 

 process of discovery was by observation rather than hypothesis. A more 

 striking example may be derived from trade cycle studies. It is an 

 accepted generalisation, not indeed possessing the universal validity of 

 the law of demand but none the less of substantial authority and interest, 

 that in the upswing of production prices have a rising tendency and in the 

 downswing a falling tendency. It may safely be said this could not be 

 deduced from the propositions of static theory nor from that part of 

 monetary theory which is deducible from them. Falling prices would 

 be regarded as an equally (if not more !) likely accompaniment of rising 

 output and vice versa. The generalisation is a direct result of observation, 

 an excellent example of the facts speaking for themselves. And if theo- 

 retical explanations have subsequently been woven round it, this must not 

 blind us to the true source of our knowledge. If rather crude observa- " 

 tional data can yield appetising morsels of this sort, may we not legiti- 

 mately hope that when subjected to refined statistical treatment they will 



