Are negative interest rates enough?

I'm going to mention Scott Sumner one more time and then give him a break.

For the last six months he's been recommending that central banks (mainly in the US, but presumably worldwide) pay a negative interest rate on commercial bank reserves.

(for reference, here's another in a string of recent articles justifying why the Federal Reserve pays positive interest on reserves, without even mentioning the negative interest option. Are they under instruction not to discuss it?)

In short, the argument is this. Quantitative easing works by paying banks cash in return for their government bonds. Since they don't earn any interest on cash (unlike on the bonds they were holding previously) they will want to lend out the newly acquired cash into the private sector so that they at least make some money on it. This (depending on your model) will boost the velocity of money, or increase the relevant measures of money supply, or increase inflation expectations, or increase investment, all of which boost GDP.

But at present, when the banks get this money, the Fed pays them interest to keep it on deposit with them. This means they have less incentive to lend it out. Scott wants to charge them interest for the privilege of holding the cash safely in the Federal Reserve vaults. This will make it more expensive for them to hold cash and (relatively) more profitable to lend the money.

Now clearly this will make some difference at the margin. For each interest rate, there is undoubtedly an optimal balance for banks to keep in reserve. If the rate is lowered by one percentage point (from 0.5% to -0.5%) this optimal balance will change.

However, will it change very much? Rational theory implies that the change from 0.5% to -0.5% is pretty much identical to a change from, say, 5% to 4%. But banks are still keeping $800 billion in reserves while they can earn (at least in the UK) up to 12% on business loans or 18% on personal loans. In economic terms, what is the price elasticity of supply on lending?

When I first saw Scott's argument, I wondered if it was based more on psychological expectations. If a bank expects to lose 6% on a certain loan (one which is presumably some way along the $800 billion curve away from the marginal decision), it shouldn't make any difference whether they make 0.5% or lose 0.5% by keeping the money on deposit in central bank reserves. So it feels highly unlikely that a negative interest rate would make a big impact on that $800 billion. Maybe $50 or $100 billion would head out the door, but the majority of the money can't be that sensitive to such a relatively small tweak.

But if the bank's shareholders suffer from loss aversion when their capital is being eroded, maybe they'd rather tell the execs to lend out the money and take a chance that the 6% losses won't really happen. This kind of behaviour is clearly observable in lab experiments; but we would like to believe sophisticated bank executives are immune to it. Are they, though?