Julian, 
 
Kevin is quite right. Crucial concepts in his answer are 'relatively' and  
'theory-independent', when he says:  
"data [regarded] as observable when it can be measured or collected in  
a relatively theory-independent  manner". 
 
The logical empricists of the Vienna circle thought that observation  
terms could be separated straightforwardly from non-obeservation  
terms. The latter had either to be explained in observation terms or, if  
this turned out to be impossible, dismissed as 'metaphysical'and, by  
implication, 'useless'. 
 
Popper had understood that all terms are in some way laden with  
theory, if 'theory' is allowed to having a rather vague meaning here. We  
now know that terms are only relatively observational. A term X can be  
laden with the very theory X* it figures in, in the sense that X* explains= 
  
X: in that case X is X* laden, hence X* relevant, hence X* guided. The  
concept of 'equilibrium price' in the theory of (subjective or objective)  
value is, I think, a good candidate. So is 'weight' in Newton mechanics. 
 
So terms are theoretical (referring to a non-observable phenomenon)  
only in a relative sense: relative to a particular theory. If there were a= 
n  
absolute demarcation between observable and non-observable, it  
would be difficult to draw the line: is what I see with the aid of a  
microscope only an observable? And what about my spectacles?  
Optical telescopes? Radiotelescopes? 
 
It is however also possible that a term Y (used in theory Y*) is not laden= 
  
wit theory Y*. In that case, Y may be Y* relevant or not Y* relevant. Y is= 
  
Y* relevant if theory Y* is not indifferent to a fact, the description of  
which contains term Y. For instance, if  Y* explains or contradicts the  
fact. 
In case, then, Z is theory Z* relevant, there are again two possibilities.= 
 Z  
may be Z* guided or not. Term Z is Z* guided if Z* is an observation  
theory and if Z figures in the description of a fact, observable only by  
using observation theory Z*. 
 
The methodology concerning observables and theoretical terms is  
much more fine grained, and more complex than this. But this is a quick  
overview of the issues concerned. 
 
As to M=E4ki, his claim is that economics uses primarily terms borrowed  
from everyday life, such as 'consumption' or 'utility'. Therefore  
economics allegedly makes no use of unobservables. From the above,  
it is clear that I disagree. The fact that the economic environment  
consists of objects, properties or relations of human measure (and not  
of the size of, say, a neutrino) does not imply that seeing them is  
always easy. To see is to intervene and the intervention may be  
possible only by the use of theories. 
 
You ask for the locus classicus for the concept of observable, but I am  
afraid that the only loci of this sort are to be found in natural science.= 
  
For economics, the debate on it is fresh: we shall have to find it out  
ourselves. 
 
Menno Rol 
University of Groningen