What do you think drives anticipations of the future, if not the recent past? There's nothing to make people think the price of an asset at time t+1 will be greater than the price at time t than the fact that the price at time t is greater than that at t-1, and t-1>t-2, etc.
[WS:] Not necessarily. Think, for example, of a piece of "railroad property" i.e. land bought in anticipation of railroad being built next to it (which happened a lot in this country's past.) Immediately prior to the transaction, the value of the land is low which reflects the demand situation - few people want to buy land in the middle of nowhere - assuming, of course that nobody but railroad officials and a handful of speculators know about the plans to build the railroad.
In this situation, a speculator may offer a price that is higher than the prevailing market price to entice the owner to a quick sale, before the news of the railroad extension break out and the price goes up. Or imagine another situation - a speculator who owns several lots of land in an area sells one to an accomplice at an artificially inflated price to convey an illusion to other speculators that a railroad will be build there. Other speculators move in and buy his other lots at inflated prices.
In each of these cases, the price has nothing to do with the current or past demand for- or market prices of land in this particular area, and everything to do with the anticipated demand, based on current privileged knowledge, be it true or falsified. That is, when the *anticipated* demand goes up, so does the price. Q.E.D.
I would like to add that I am not trying to defend NCE, but rather think of an argument that would save the NCE's core assumption - the price-demand relationship - against empirical evidence. Introducing an unobserved but inferred factor (expectation) does the trick, and AFIK NCE hacks use that trick a lot. The "only" drawback of this approach is that it transforms NCE statements from empirical propositions into tautologies true by definition.
Wojtek