F— ECONOMIC SCIENCE AND STATISTICS 159 



yield more fruit in plenty ? It will still be necessary to relate such 

 generalisations to each other and to those of a more deductive origin in an 

 orderly fashion. 



Having made this plea for the more radical empiricist, I will conclude 

 by mentioning one or two types of investigation suggested by the present 

 condition of theory. If I make no mention of others now under way, I 

 hope it will be understood that this is not because I regard them as unim- 

 portant, but for lack of space and because the former happen to have 

 caught the speaker's eye first. 



Emphasis has recently quite properly been placed upon the importance 

 of expectations with regard to the future in determining the present 

 actions of the individual, and upon the slender basis of knowledge on 

 which he is obliged to form his expectations. Speculation upon the 

 consequences of this may therefore be regarded as arising directly out of 

 theoretical considerations. 



Ignorance with regard to the future drives the agent back to an im- 

 perfectly rational dependence upon past experience, particularly his most 

 recent experience. It is reasonable on this basis to make the hypothesis 

 of a time-lag between certain adjustments. By introducing a systematic 

 lag it is possible to give a mathematical demonstration that an oscillation 

 of behaviour must result. The interesting survey by Dr. Tinbergen 

 (1935) in Econometrica discusses a number of hypotheses of this 

 nature. 



Statistical verification may proceed from two ends. On the one hand 

 it may be possible to verify the particular lag assumed by reference to two 

 statistical series. On the other the cycle mathematically deducible from 

 the assumption of such a lag may be compared as to its general features 

 with the real cycle. One might hope that even with the data already 

 available the determination of lags in this empirical manner might give us 

 a theory of the trade cycle, which would be self-consistent and consistent 

 with the broader generalisations of theory and also subject to fairly 

 approximate empirical verification at both ends. Fortified by such tests, 

 with what far higher degree of confidence might we call upon legislatures 

 to take remedial measures ! I may add that the framework of equations 

 within which the lag hypothesis should be applied are those of dynamic 

 economics. This gives another reason for wishing an early precise 

 formulation of these. 



I now pass to an entirely different type of empirical work. General 

 considerations suggest that the entrepreneur acts under the influence of 

 certain defined forces. When we come to examine these, it is surprising 

 how largely the entrepreneur must be ignorant of their precise value. This 

 is evident enough in the case of capital outlay, decisions regarding which 

 must be based on prognostication. But even current output is properly 

 determined by reference to the value of the loss or gain of customer good- 

 will and to that of ' user cost ' (Keynes), both of which depend upon 

 prognostication. And apart from the future, there are other matters of 

 uncertainty. Correct behaviour in the field of imperfect competition, and 

 this is the greater part of the whole field, presupposes knowledge of the 

 value of marginal revenue, which in its turn requires knowledge of the 



