Given the choice of too-big/too-little and the consequences, I'd lean toward too-big. Not paranoid here.
Corruptions or failures in big organizations -- and that doesn't matter whether it's the government, business, or any other organization -- have greater negative consequences.
We also know all organizations are imperfect, it takes a lot of time and ethics to keep them from becoming selfish.
The Wall Street situation wasn't simply a failure of big government, it was also that too big businesses had been allowed to exist. Relaxing business regulations at a time we simultaneously relaxed anti-trust enforcement AND consolidated oversight in fewer agencies is a bad combination.
You're better off placing many small bets, accepting many small losses and many small gains, then one big bet on the assumption that you know all the possibilities and can make a statistically valid decision. Nicolas Talib's book Black Swan is a great one on this topic -- debunking the idea of the bell curve and trying to manage most things using conventional statistical methods.
So you're better off having many smaller businesses, and multiple layers of regulation (hey Federalism, what a concept!)
As it applies here, you're better off pushing as much decision making -- along with the research and broad policies to help guide them -- down to the lowest practical level, such as a ranger district or maybe the whatever level is above them.
You still provide oversight through audits and observation to make sure there isn't malicious or simply too greedy activities.
What will happen when you devolve decision making to smaller units is you will have more failures, but those failures will be contained and relatively small impact. It also gives you the ability to point to districts that things are functioning well and go, "That's what we should be trying to be like!"
When you consolidate decision making to bigger units in government you may nominally have fewer "failures" but when you do they're doozies and have great negative impact due to their size. And you lose the ability to compare the performance of A to B and figure out which is the better way to do things.
