Firstly I need to know at what level do you consider the system secure enough for transactions: so please reply with your minimium average cost to an attacker for a successful attack for the values of $1000, $100,000 and $10,000,000US.
(eg: for me personally I would feel comfortable knowing that it would cost for a small transaction <$1000US atleast 2x that to mount a successful double spend attack (ie: if the attacker spent twice as much as they stole they it's good enough for me), if for a moderate individual sum say <$100,000 I'd like feel comfortable that it cost atleast 5x and for a large sum <$10,000,000 say 20x
(please note: just edited 1x to 2x for under 1000US)
All that is required, IMHO, is that the cost of a double spend attack to likely be higher than the potential ill-gotten gains. So if you go and buy a car for $50K in BTC, it should be more expensive an attack than that. However, you are already overlooking a feature of Bitcoin, the more 'confirmation' blocks that one waits for after the first one that accepts your transaction before you walk away, the more "expensive" (in several ways) that such a double spend attack becomes. Also, it becomes more expensive at an exponential rate. Thus, anyone who is selling you that car, considering the sums involved, is going to want to wait for 6 confirmations or so before letting you drive away. At the current hashrate of about 400 PetaFLOPS equivalent, it would take at least 20 of these to overcome just
one block confirmation...
http://en.wikipedia.org/wiki/Titan_(supercomputer)
Therefore, the current level of security is several orders of magnitude beyond what is necessary to disincentivize a frausdster from even attempting a double spend attack. We crossed that point around 2010.
The security that is being paid for is to protect teh entire system from an institutional attack on the blockchain itself, and there is (likely) not a single nation-state with the spare resources to attempt it for even a few hours. So I'm going to assume that the current profitablility for miners is more than sufficient to secure the blockchain.
Thus the question then becomes, how do we make sure that the curernt level of profitablity continues after the block subsidy is reduced? First off, that may not realy matter for decades, as teh growth in value has thus far outweighed the reduction in block subsidy. What was the exchange value when the subisdy dropped from 50 BTC to 25? I know that it was under $15, and I'm fairly sure that it was under $10. So at the current price of $34, we are already well over double the profitablity for miners overall. As long as the real spending value of a BTC continues to double within a four year period, the concern is moot.
But, of course, eventually it wont. Such growth is not sustainable, so at some point the value must stablize. How, then, can we be certain that transaction fees will be enough for miners to continue to secure the network?
In short, we can't really know this, but the economics of the system imply that we can expect that an equilibrium of fees will be reached in one fashion or another. So long as the overall Bitcoin economy is large & mature enough by that time, a tiny fraction of the GDP would be required to incentivize miners into the foreseeable future. Far less, in fact, than what is taken from you via inflaiton of fiat currencies; which are
at least 2% of their entire monetary base every year. The big key is that Bitcoin is much more economicly efficient than fiat currencies are.