1 
PART II 
INTRODUCTION 
The following papers on stochastic processes have been 
separated from the others because their prospective readers are 
presumably a separate class, too. I do not like it this way, 
though. I would rather like to bridge the gap between the two 
subsets of my readers and to convey to them the fundamental 
unity of purpose and outlook of my endeavours. 
Why do I use an approach which offers such overwhelming 
technical difficulties and yet requires, if anything,even more 
simplification than ordinary deterministic models? My aim in 
this type of work is a seemingly modest one:I want the 
economist to take a new look, with different eyes,as it were, 
at the statistical material which so far they have very often 
interpreted with a disarming naivete. Perhaps I may give an 
example. Some economists- a good many years ago, it is true, 
have been puzzled by the fact that time series and cross 
section data give different results when they are used, for 
example, to estimate the propensity to consume. They put 
forward all kinds of fancy explanations but they did not ask 
the obvious question why an identity of these two kinds of 
measurement should be expected at all. They might have tried to 
link the question to a closely related problem in statistical 
mechanics: The ergodic theorem which deals with the equality of 
"phase averages’* and ’’time averages” ; this theorem is valid only