Liquidity and IOR

Re: the big balance sheet and how it improves financial stability.

Rodney Garratt, Antoine Martin, and James McAndrews at the New York Fed have a very nice post, Turnover in Fedwire Funds Has Dropped Considerably since the Crisis, but It’s Okay.

Before the crisis, banks held about $50 billion of reserves at the Fed. That's not a lot of money. When banks want to pay each other -- say you write a check to me, so my bank has to get money from your bank -- they do it by transferring reserves through the Fedwire.  So, that's why banks keep some reserves there.

But $50 billion is tiny compared to $10 trillion of M2, and banks use reserves to clear financial transactions too. A huge amount must flow by passing around these tiny reserves. How did banks do it? What happens if bank B says to bank A, "send us $10 million" and bank A didn't have $10 million left at that second in reserves?

Answer: "intraday overdrafts." The Fed would lend bank A the $10 million -- just flip a switch and put $10 million in their reserve account, and call the loan an asset corresponding to this liability. A then pays B, and works hard to make sure that it collects $10 million from C and D by the end of the day.

Source: Rodney Garratt, Antoine Martin, and James McAndrews at the New York Federal Reserve

As you can see, such "overdrafts" accounted for 50-60 percent of all Fedwire transactions before the vast increase in reserves.

It's a system that makes a lot of sense, so long as banks never fail and don't abuse it. It allows the system to produce a much higher volume of transactions with less non-interest-bearing assets. Instead of cash in advance for every purchase, settling up once per day means you only need to cover the worst possible daily total flow, not the worst possible intraday flow, like if $10 million goes out 10 minutes before another $10 million comes in.

But now, banks have $4 trillion of reserves. They're sitting around as investments, really. As long as they pay full market interest, there is no reason for banks to go to all this effort to get by with little reserves. And we seen in the graph exactly what you'd expect. If bank A owes bank B $10 million, it just sends the $10 million, no need to borrow it for 10 minutes from the Fed.

The article explains all this well. A few quibbles

The shift in funding away from overdrafts and toward account balances has significantly increased the amount of liquidity needed to fund payments in Fedwire Funds. 

I think reality is the other way. The vast amount of liquidity banks have chosen, and will continue to choose so long as reserves pay market interests, mean they have abundant liquidity to fund payments directly on Fedwire. It is not "needed." (Mistaking "choice" for "need" is a favorite Econ 101 fallacy.) The minute the Fed tries to pay less on reserves than short term T bills pay, banks will choose to go back to the old system.

And turnover -- which they point out has plummeted as in the graph below (ignore the "counterfactual") -- is a totally misleading statistic. Turnover is transactions / reserves. Transactions haven't fallen, reserves have exploded. I presume a graph of the total number of transactions shows little change, or at least no such cliff.

Source: Rodney Garratt, Antoine Martin, and James McAndrews at the New York Federal Reserve

But the closing paragraph is great:

A high turnover ratio is typically viewed as a good thing in a payment system, because more payments can be made with less liquidity. To do more with less is good when resources are scarce. However, reserves don’t have to be scarce. With interest on reserves, the Fed can implement monetary policy even though banks are flush with cash (as we noted in this Economic Policy Review article). And because banks have less need to economize on liquidity, payments are made earlier in the day, which benefits consumers and increases the resiliency of the system to operational outages or participant failures. So the large decrease in turnover should be viewed as a good thing; it is another symptom of how the high level of reserves benefits the payment system.

"Payments are made earlier in the day" is important. Demands for payment earlier and earlier in the day are a key part of failures.

H/T to Torsten Slok's weekend reading email which found the post.

Update: "Interest on Reserves and Daylight Credit" bv Huberto M. Ennis and John A. Weinberg in the Richmond Fed Economic Quarterly (2007) is a nice explanation of how the system worked. Towards the end it sketches how increasing reserves drive lower turnover, not less transactions.