By Jove! Someone’s finally got it.
Cutting interest on excess reserve is a hugely risky option for the Fed, and could do more damage than good (leading even to major systemic issues). We’ve said as much, and now RBC Capital markets makes the same argument too. But much more eloquently (dare we say).
The main reason, of course, is that cutting IOER could wreak untold havoc in the money and repo markets.
It’s all inter-related to things like the Treasury fail penalty brought in May 2009 and the FDIC fee, which brought in in April 2011 (about the same time Bill Gross of Pimco started talking down and “shorting” US Treasuries, a fact which eased repo market stresses — an interesting coincidence).
As RBC Capital Market’s Michael Cloherty and Dan Grubert explain, all of these factors have led to a situation where Treasuries can now trade at a negative rate. Something not apparent for other markets (yet) because of there being no penalty systems in place there. In other markets if you come close to zero, you just fail instead:
There is a 0% floor on repo for everything except Treasuries, because the effective cost of failing to deliver a security is equivalent to borrowing that security at 0%. The floor on Treasury repo is -3% due to a 300bp fee for failing to deliver a Treasury. The repo floor puts a limit on how special specific securities (ex-Treasuries) can trade when general collateral rates fall: when GC is at 20bps a security can trade up to 20bps special, but if GC falls to 5bps that issue can only trade a maximum of 5bps special.
This is the key reason why after the Lehman crisis (before the Treasury fail charge was brought in) there was a rampant spike in Treasury settlement fails, and why after a fail charge was brought into that market, the fails mutated elsewhere.
Of course, as we’ve noted, there’s recently been a huge spell of negative repo trading — mainly because of the cornering of one particular benchmark issue, which has seen that security trade as low as minus 2.92 per cent, and before that because of the introduction of the FDIC fee. All of which is having an impact on all Treasury repo rates, and even General Collateral (GC) rates too.
And herein lies the risk:
As GC approaches the floor, at some stage the spread that bondholders can earn on a securities lending trade becomes too small to justify the credit risk, operational complexities, and transactions costs. At that point bondholders stop lending and it becomes extremely difficult to borrow bonds to make delivery on a trade. As the following charts indicate, historically low overnight rates have tended to sharply increase the amount of fails in the system, as those low rates limit the incentive for bondholder to lend their collateral.
The classic example of the rate impact on fails was in the wake of the Lehman bankruptcy, when overnight Treasury GC fell to the low single digits and there was no fails fee to allow repo to trade negative. Many bondholders stopped lending, and fails went to record highs. When the 300bp fails fee on Treasuries was introduced in Q2 2009, it pushed down the floor on repo, boosted the incentive for bondholders to lend, and sharply reduced the amount of fails in the system. If lower IOER causes GC rates for non-Treasury repo to fall too low, then we are likely to see enough bondholders pull back on their lending to cause very large systemic fails.
In a nutshell, more fails means more residual credit risk between the two counterparties in a trade.
Or as RBC explains it:
If a buyer purchases a security at $102, the seller fails to deliver, the security rallies to $108, and the seller goes bankrupt prior to delivery, the buyer loses the $6 appreciation in the security. Meanwhile, the seller is at risk if the price of the security falls and the buyer goes bankrupt prior to delivery.
Under normal conditions, this risk is not meaningful because fails usually clear up quickly. This means less time for prices to move substantially away from the initial market price (a smaller loss given default), and less time for counterparties to become insolvent.
But if there are systemic issues that cause persistently high fails, then the odds of a large loss due to a fail rises sharply.
In the extreme fails scenario that extremely low GC rates could cause, investors would need to incorporate the potential for fails-driven counterparty exposure in bid/offers. This means different bid/offers for small hedge funds, large hedge funds, weaker banks, stronger banks, etc. This would be completely at odds with the “one price” theory behind Trace reporting. And it means that investors will need to think about credit limits to their dealer counterparties before they trade cash bonds, complicating trading and hurting liquidity in the cash markets.
But how does all this relate to IOER? Well simply because IOER acts a floor on rates, helping to keep them supported. (It’s also a form of sterilisation for the high-powered money that’s created via quantitative easing.) If you can receive 25bps at the Fed, the Fed Funds effective rate itself will never be influenced by negative repo rates elsewhere. It won’t go negative. Remove IOER or lower it, however, and there’s suddenly a genuine risk that even zero can’t be protected.
Of course, because failing on Treasuries now carries a 3 per cent penalty fee, the IOER rate would arguablly have to be as low as minus 2.5 per cent before it really started to impact Treasury GC. Thus the immediate impact on of a cut on IOER would really be on the regular GC collateral rate.
But what would that mean?
As we’ve noted, the introduction of the FDIC fee in April eliminated a very well exploited arbitrage for banks, which saw them borrowing cheaper than 25bps (the IOER rate) from Agencies, who did not have access to IOER, and parking the cash at 25bps at the Fed. The difference was a risk-free profit, and helped to keep the Fed funds in and around 25bps. The FDIC fee ate into those profits, however, discouraging banks from conducting the trade — a fact which immediately put negative pressure on repo rates.
The FDIC fee, however, doesn’t apply to foreign banks, which means they are the main ones left exploiting the IOER arbitrage. Cut the IOER, and that kills that trade, says RBC:
…we think the tie between IOER and repo on spread product will operate through a large flow of reserves from foreign banks to US banks. Foreign banks now have roughly $850bn of reserves at the Fed. There are three reasons they hold these reserves: first, to have a dollar liquidity cushion; second, to meet regulatory liquidity requirements; and third, because they can borrow at rates below 25bps and leave their cash at the Fed to earn 25bps. Note that US banks are limited in their arbitrage of IOER because they must pay an FDIC fee on total assets.
Where there is no data source that can tell us the size of this arbitrage flow by foreign banks, we do note that when RP rates rose sharply for a few days due to the debt ceiling, foreign reserve balances fell by $132bn in two weeks and reserve balances of domestic banks rose by $125bn (reserves are a zero-sum game). This large reception to a brief move in rates suggests that the arbitrage flow is quite large—several hundred billion dollars.
If IOER fell, then foreign banks would no longer be able to capture a spread by borrowing money and leaving it at the Fed. These banks would cut back on their dollar borrowing, meaning fewer CDs, less CP, and less lending of collateral in the repo market. This drop would exaggerate the supply shortage facing money market funds, causing all short money market product to richen. GC for mortgages, corporates, etc would richen, driving them close to the 0% floor.
Note that the resulting large reserve flow back to US banks would heighten the balance sheet pressure at many institutions. That pressure would also be more intense because lower IOER means that banks would lose even more money on their reserve balances—currently reserves pay 25bps but banks must pay an FDIC fee on the order of 15bps. In addition, they must fund reserves the same way they fund any other asset, and the marginal cost of funds is far in excess of 10bps. Accordingly, the banking system is already losing money on reserves with IOER at 25bps—a cut in that rate would mean putting the banking system into an even larger $1.6T negative carry trade.
Which means in effect a lower IOER would put more balance sheet pressure on banks (overload them with more reserves) and make it increasingly difficult to make competitvely priced loans, meaning net interest income would come under pressure. The only solution would be to pass those costs along to customers through things like negative interest rates on certain deposits.
To think of it another way, it would introduce a cost on money.
Which of course is possibly what the Fed wants to do to encourage the money to flow into the real economy — but it also runs the risk of kicking off a deflationary spiral that will be impossible to stop, especially if market expectations catch up with Fed thinking as regards deflation risks.
As Lars Svensson, deputy governor of Sweden’s Riksbank, noted in a paper in December 2003:
Nominal interest rates cannot fall below zero, since potential lenders would then hold cash rather than lend at negative interest rates. This is the socalled “zero lower bound for interest rates.”
The problem is that the economy is then satiated with liquidity and the private sector is effectively indifferent between holding zero-interest-rate Treasury bills and money.
In other words, money loses the unique characteristic that makes it the optimum liquid instrument for exchange. It loses velocity and instead becomes a store of value.
People don’t differentiate between it and zero-interest-rate Treasury bills.
Vaults of the stuff would accumulate, leading to legislation possibility prohibiting stockpiling or a ban on reserve convertibility into banknotes (a ban which never happened during the Great Depression, and which some say made the crisis worse).
The perils of releasing the repo rate – FT Alphaville
Negative interest in cash, or goodbye banknotes – FT Alphaville
Fantasy Fed Options – FT Alphaville
When a government bond becomes a Giffen good – FT Alphaville