I prefer to give the original date in citations where possible. But 
it isn't always easy. What about things that were not published at 
the time (e.g. Smith's lectures), or modern collections with items of 
varying original dates (collected economic works of ...)?
Incidentally: Charles McCann complains that the Oxford (=Glasgow) 
edition of the Wealth of Nations is 1789, not 1776. But it is a 
variorum edition which (using the footnotes) is 1776, or 1789, or any 
edition in between. So it is implicitly a collection of items of 
varying dates of the sort mentioned above. The best answer in these 
cases may be to mention the date of the particular item under 
discussion in your main text, not just the references, where it is 
relevant, for example in discussing additions to the Theory of Moral 
Sentiments in the last 1790 edition.

Tony Brewer