Like ArticMine pointed out, there is already a cost to produce a large spam block, regardless of whether the costs is paid by a user in the form of fees, or by a miner in the form of increased orphan risk.



Smooth is also correct, however, that just because the spam block has a finite cost and that a fee market exists, it doesn't necessarily mean that the fees will be the "right" fees to keep people who want to run Bitcoin nodes on low-cost hardware with slow internet connections happy, for example.  As the chart above shows, both transaction fees and spam costs (measured in bitcoins) decrease as the propagation impedance of the network improves.  

This relates to the question of externalities.

(a) Does retaining the block size limit as an anti-spam measure (i.e., a production quota above the free market production Q*) result in a negative externality?  If so, is it worth it to implement a production quota below Q* and suffer a loss of economic activity equal to the region shaded in brown?

(b) If the answer to both questions in (a) is "yes," then the question is whether it is even possible to enforce a production quota without a strong organization or government willing to use force if necessary.