This is a draft of something I've been thinking about for the last week. Unfortunately, I don't have time to look at it for a minute longer, so I'm posting it as is. please forgive the slightly less-than-perfect organization
There seem to me to be two fundamental ways of protecting the financial system. The first is to restrict what financial institutions can do (let's call this "activity fencing") and the second is to restrict the amount they can do it (which is basically deleveraging).
Activity fencing has been a common cry, but I suspect it is far inferior as a method. A number of reasons come to mind.
Firstly, most of the activities that banks would be restricted from doing are services that banks provide to clients to assist them in reducing their own risk. For example, much of the widely-disparaged prop trading is the bank helping a client de-risk and holding the other side of the deal on its own balance sheet. In many, though not all, cases, this is simply until the bank can find an appropriate counterparty.
Hedging in this manner can be a big win-win for clients. Think quickly about upside-downside risk of an airline and a refinery hedging the price of jet fuel. Both firms, highly capital intensive and often themselves quite levered, face far more existential risk from an unfavorable move in jet fuel than they benefit from a favorable one. Hedging is by no means zero sum in impact, even if it is in cash flow.
Thus, banks providing these services is a form of risk pooling, no different from and in many ways superior to that found in a health insurance company. If we aggregate all the risks together, many will cancel out, so that the net remainder of risk is far less. A lot of this remainder ends up on financial institution balance sheets as prop trades.
Another analogy is inventory management. It is fundamental in operations management that the total amount of inventory you need to hold is lower for a pool of inventory serving lots of outlets than it would need to be on an aggregate basis if each outlet kept its own inventory. This is because each outlet risks over- or underpurchasing relative to inventory levels but must keep a safety stock of extra inventory to handle high demand periods. Aggregating up reduces the standard deviation in demand (outlets that are oversubscribed can cancel out those that are undersubscribed) and allows that safety stock to be significantly smaller. We can think of risk at banks in a similar fashion.
The true problem of these activities, which legitimately reduce risk on their own, is that it is not easy to measure how much you have actually reduced risk, and if you overestimate how much risk you've reduced, you take on too much additional risk and end up with an institution that is too risky. I hope it is clear in this framework that the issue is not one of "useless" activity but instead one of measurement of just how much that activity is helping.
Addressing this fundamentally must happen by restricting legitimately helpful activity where it is very hard to measure the amount of help it provides on the margin. The question is whether this should be done more by restricting types of activity (activity fencing) or restricting amounts of activity (leverage).
Restricting types of activity leaves a lot of risk to corporations. Corporations are generally not too bad at knowing the risks they're going to face, and they tend to be pretty efficient in identifying the risks they want to fix. Restricting bank activities thus brings down significantly the corporate activity level by limiting them to the risk they can handle unhedged. This sounds nice, but insurance would get significantly more expensive and harder to get, anything commodities-based (airplane flights, food, etc) would be more expensive and harder to get, etc.
In fact, thinking about returns to scale, it is very likely that the first credit in a hard-to-measure activity is still a much better one to have societally than the marginal (last) credit in an easy-to-measure one. In fact, that marginal credit in an easy-to-measure area is still probably hard to measure when levered up. Would you rather hedge Berkshire Hathaway's fixed/floating interest with a swap (hard to measure but clearly safe) or a subprime mortgage (much easier to measure but not deterministically so, and risky!)
Given this, it seems far more logical to focus on deleveraging, because by deleveraging you prevent banks from taking on too much additional risk regardless of whether they think they can handle it on a measurement basis. It also eliminates the marginal credits that would be the sources of the most problems in a given financial product if that product were to get hit.
Now, it's important to note here that banks need a return on equity capital to make them able to retain investors. This return may go down as risk goes down, but it certainly does not go down proportionate to risk decreases because of the opportunity cost of capital for investors.
Given this, in the long term, significantly deleveraging banks would almost certainly reduce credit availability and increase interest rates for those who can get credit. It would necessitate a substantial increase in the money supply to maintain systemic liquidity. Note that we have already observed deleveraging, reduced credit availability and a flood of new money. Interest rates to individuals or organizations who aren't great credits will almost certainly have to rise from here - banks are almost universally earning an ROE below any decent measure of their cost of equity capital, and while it takes time for them to rework their balance sheet, a delevered bank needs a lot of equity and this will eventually revert upwards, which means lousy depositor rates continue but creditor rates creep upward.
The dynamic leads us to the unfortunate conclusion that it is close to impossible to have credit easily available to groups with credit risk (small businesses and low income individuals) while also having a highly stable financial system. Countries that have avoided this problem are few, far between and largely riding a commodities boom out of China that won't last forever. I'm trying to think of adjustments to the tax system or effective government programs that could mitigate this credit availability issue but all of the solutions I can think of are highly inefficient.
This also makes me think about the efficacy of monetary policy under different leverage levels. Safer credits are safe because they have predictable businesses. Adjusting interest rates may not affect their borrowing and investing decisions as much as it would for lesser credits and we may thus need more aggressive monetary policy responses for the same effect under a safer banking system.