Let's see some illustrative calculations:
Let's assume bASIC 57GH/s uses 120 W.
The cost of electricity (say 0.11 $/kWh) is about $58 per year higher than that of BFL SC Single.
If one buy this instead of BFL SC Single, the price difference ($230) pays for the more power for 4 years.
Let's assume bASIC 57GH/s uses 180 W.
The cost of electricity (say 0.11 $/kWh) is about $115 per year higher than that of BFL SC Single.
If one buy this instead of BFL SC Single, the price difference ($230) pays for the more power for 2 years
Let's assume the difficulty is 10 times greater than now (10 x 3054627), one block gives 25 BTC, and 1 BTC is $12.61.
In unlikely case of being them constant for a year, the Single net annual profit is 4488 $/year and bASIC net annual profit is 4261 $/year.
The difference is... $227. Interesting coincidence. (I didn't set the BTC price, it was left in my BTC calculator from two weeks ago.)
For 100 times greater difficulty the yearly net profits difference is $396 - $374 = $22.
I believe the differences are way below an error coming from uncertainty (of tech specs, of difficulty changes, of price changes, etc.).
The date when the mining starts is surely the most important factor.
This is a good post. When the power cost is small compared to the hardware cost, it becomes a negligible consideracion. In GPU mining power cost was important because one would expect to pay several times the electricity over the cost of a graphic card, especially in the case of cheaper GPUs. Obviously these untra efficient Asics are very pricey compared to how much one is going to end up paying for electricity so unless bAsics consume 10 times the wattage of a BFL Single this is not going to be an issue. More important is the release date and price per GH as others have said. Another considerations above power usage is modularity and ease of set up, and of course custormer service and miner support.