And yes, that is my point about how to account for it (and that both $120B and $120B x n are bad ways to express the loan significance, for opposite reasons).
Here I was addressing the narrower, quoted point, that they're loans "with short expiry built in". No, when the intent from the beginning is to keep rolling them over, that should no longer count as a loan with "quick expiry built in".
And if that's cheald's point, then it's invalidated. A loan of effectively three months is not lubrication, it's the [can't continue metaphor under HN rules, but you get the point].
Is there any evidence that these repo loans are being rolled over? The Fed absolutely rolls over its open market operations assets, and open repos between banks roll over, but I haven't seen anyone suggesting that term repo loans the Fed is currently engaged in are being rolled over. Lending $120b to Bank A today, and then closing the repurchase and lending $120b to Bank B tomorrow isn't evidence of systemic failure within either bank. It is evidence of systemic friction in the repo market, which is already pretty well established, and is the reason given for the liquidity provision in the first place.
It's also worth pointing out that these aren't just free loans propping up banks short on capital - they're collateralized (in many cases, over-collateralized) with treasuries. The Fed isn't "loaning the bank lunch money", it's buying the bank's valuables, which the bank agrees to buy back at a premium the next morning.
The banks should only sporadically need money from the Fed, but it is not uncommon for them to need money to cover reserve requirements. Traditionally, they purchased those shortfalls from other banks on the repo market. Post-2008, the Fed started offering interest on overnight reserves on deposit with the Fed, which gives banks with extra reserves incentive to not loan to other banks, but to simply maintain their reserves. Additionally, banks have a preference for cash reserves over treasuries, since they're easier to move around without signalling the strength (or lack thereof) of the bank's position. All that has led to a constriction of the liquidity available in the repo market, which is why the Fed is stepping in to provide that liquidity. Banks sourcing their reserve coverage from the Fed is a new development, but banks needing to cover their shortages via repos is not. The most significant change in behavior is, broadly speaking, on the supply side, not the demand side. There have been changes to the demand side - notably the increased reserve requirements post-Lehman - but it's the supply crunch (and associated rate spikes, like what happened in September) that's predominantly causing issues.
And yes, that is my point about how to account for it (and that both $120B and $120B x n are bad ways to express the loan significance, for opposite reasons).
Here I was addressing the narrower, quoted point, that they're loans "with short expiry built in". No, when the intent from the beginning is to keep rolling them over, that should no longer count as a loan with "quick expiry built in".