Running costs are extremely low for NXT: As most users have flatrates, it basically comes down to power costs. Even running the forging client on a netbook would only cost about 25€ per year.
It's extremely low, if you're only running Nxt on top of what you'd normally be doing. But remember that some folks won't be leaving their computing devices on for long periods of time. Even if you assume that e.g. Nxt node is set to automatically run whenever you connect to the internet, there are still other costs (that usually aren't factored into the numbers) like setting all this up, one more process to consider as culprit if your system hangs, etc.
As the price will increase and it is always possible to reduce power consumption by using a rasp-pi and a solar-panel
Rasp-pis may not suffice for much longer, there have been discussions about this recently, though I'm sure there will be newer, more powerful and still cheap microboards that will take their place. And solar panel production costs are going down and efficiencies are going up. But they're still a non-negligible cost in terms of $ and setup.
Basically, if I'm a regular guy (i.e. not someone who has a comp running for long periods of time, and not particularly concerned about the Nxt network) and my expected net gains are very low, then I'd rather not bother with forging at all. Too many things in life to worry about, I'd rather just lease my forging power and have one less. Let the market make it easy for me to decide where to lease it, e.g. who can offer better returns, which forger seems to be not doing evil.
EDIT: So ultimately it'll depend on price and transaction volume increasing. If they increase enough, then regular folks may find it worth it to forge themselves. Then yeah, I can see shared forging, or some other form of variance reduction in the protocol as much more appealing than leased forging or joining pools.
. 
