David Henderson noted that capital cushions are not cushions at all, highlighting a perverse effect of these requirements. One of the more interesting puzzles in banking is the absence of any clear leverage-risk nexus in the statistical data. I know everyone thinks the past recession proved leverage is an obvious and soluble problem, but that's anecdotal, and I doubt that Bear, Fannie, or AIG would have survived with twice their capital. Hyman Minsky's financial instability hypothesis remained obscure until after his death precisely because there wasn't clear evidence between leverage and business cycles (as an acolyte, I would have loved proving the big man right...alas, he missed a lot).
When I was head of capital allocations at a bank, every week I would be in a meeting or presentation where the logic of our mission was that higher 'risk' required a greater 'cushion' for bad times, a greater equity charge. I still think this is a good idea (no better alternative), but the whole 'cushion' idea simply doesn't generalize. We (consultants, regulators, advocates) looked cross-sectionally at banks, but could not see a clear cost-of-funning and leverage connection, and it would have helped me immensely to have such a finding. Also, one of the big goals in modeling default risk would be to create a financial firm model, because many banks and finance companies do not have agency ratings. Again, I worked at an institution that would have loved demonstrating this, but there is almost zero correlation between leverage and default rates.
Given the financial sector has been in aggregate insolvent in 1975, 1981, 1990, and 2008, I would say it is generally undercapitalized by a factor of two. Yet, proposals to require a capital amount of, say, 8% of assets counterproductive. This is because a good portion of a firm's value is its franchise value, the present value of its brand, and all the myriad regulatory contrivances that keep competition at bay. Now, if a firm's capital amount is 8.01%, as a debtor I realize this firm is a coin flip away from losing that franchise value, and then I might have a principal loss on my debt. The "capital cushion" becomes simply a new trigger value, as opposed to resource to tap in tough times.
The bottom line is, as long as the private sector anticipates little or no risk, it will skate close to the edge, which will be the regulatory minimum, not zero (either spell death, and thus a liquidation of sorts). Perhaps, given the politics of closing failing banks, there should be a bank asset size limit because we can anticipate a Goldman simply having too many connections to ever let it fail, whereas ten firms each with one tenth the size wouldn't have such pull. I can be convinced of that. As per the private sector underestimating risk, I agree this happens, but only after a couple generations. The safety of the 1950s was not due to regulation, it was because these bankers had experienced failure first hand and such lessons leave a mark (did any actually read the SEC act of 1933 and 34? It simply has brokers sign statements saying they won't rob you blind, an oath administered by Joe P. Kennedy). Just look at how Commercial Real Estate has done subsequent to its 1990 crisis (lower-than-average defaults), or how tech firms did after their 2001 debacle, to see that while the public is fooled again and again it's not the same generation on the same products.
So, bankers and everyone else believed there was too little risk in mortgage lending. Currently mortgage debt needs 50% down payments to get securitized in the private market, so they don't need top-down reminders that this is risky. Indeed, the government is lending at 3.5% down (FHA), so I would say they aren't exactly leading by example.