Basel III’s Net Stable Funding Ratio (NSFR) was described in a previous blog. This piece identifies some of the undesirable impacts of the proposal.
Unsecured versus secured funding
One of most notable features of the NSFR is that unsecured and secured funding sources are given the same ASF factors. In the case of interbank funding of six months or less, unsecured or secured, this is 0%. In other words, short-term secured funding, such as repos and other SFTs, is deemed to be completely unstable.
In the case of funding beyond six months, unsecured funding is treated more favourably than secured funding. This is because the collateral on the balance sheet of a secured borrower such as a repo seller is regarded as ‘encumbered’ (unless the repo is with a central bank) and is subject to enhanced RSF factors. Where the period of encumbrance (the term of the repo) is for more than six months but less than one year, assets which otherwise would have RSF factors of 50% or less, are assigned RSF factors of 50%. Encumbrance of more than one year pushes the RSF factor to 100%. In contrast, if one used unsecured funding to buy the same assets, they would not be regarded as encumbered and so would retain their original (lower) RSF factors.
It could have been worse. The 2010 NSFR proposed to give unsecured corporate funding of less than one year a 50% ASF factor but impose a 0% factor on secured funding of between six months and one year. However, the 2014 NSFR proposes that both types of funding should have a 50% ASF factor.
This is all illogical. Repo against HQLA proved a resilient source of funding during the recent crisis and far more robust and with much less roll-over risk than unsecured alternatives. And the supervisory definition of an HQLA is based on proven repo market liquidity in times of stress. Moreover, the actual encumbrance of repo collateral is not 100%. It is fairy marginal (due to haircuts or initial margins and possible margin calls).
Basel would seem to harbour a suspicion that, contrary to the evidence, all short-term repo is an unstable source of funding for ‘shadow banks’. This is possibly driven by a US bias which reflects the peculiar instabilities of their repo market and, in particular, the dominant tri-party repo segment.
On the lending side, the NSFR also applies the same RSF factor to an asset, whether it is purchased as collateral in a reverse repo or purchased with unsecured funding. RSF factors are based only on the type of counterparty and the term of lending, and are applied to the market value of the collateral. This ignores the fact that an asset taken as collateral in a reverse repo may include a haircut or initial margin. Indeed, the RSF factor duplicates the haircut or initial margin.
RSF factors versus haircuts/initial margins
RSF factors and haircuts/initial margins do the same job. RSF factors are ‘intended to approximate the amount of a particular asset that could not be monetised through sale or use as collateral in a secured borrowing on an extended basis during a liquidity event lasting one year’. As noted already, this means that RSF factors duplicate haircuts/initial margins in the case of repos and equivalent securities financing transactions (SFTs). In addition, there are doubts over their size. Because the NSFR is intended to be simple, RSF factors are typically at the upper bounds of recognised levels of haircut/initial margin. Whereas the RSF for government bonds with a Basel II risk weight of zero is 5%, the ECB charges 0.5% to 5.0% for AAA to A- bonds, depends on the remaining term to maturity. LCH-Clearnet SA charge 0.2% to 3.5% for French government securities with no more than 10 years to maturity (collateral is typically below 10 years to maturity because of the higher volatility of ultra long bonds).
Do RSF factors have to be higher because they apply to a stress period of at least one year and the market value of an asset can fall further over one year than the typical terms for which repo haircuts/initial margins are calculated? This argument ignores the margining process under master agreements like the GMRA and GMSLA which aims to ensure that haircuts/initial margins are maintained. And if the liquidity of an HQLA can survive multiple market-wide stress events over 30 days, why should it suffer due to firm-specific stress over a year?
RSF factors are, in part, intended to reflect the assumption that banks will roll over lending to customers to preserve relationships. However, this is an unrealistic assumption in the case of securities, where there is no relationship with the issuer.
It is reported that some regulatory authorities wanted an RSF factor of 100% for HQLAs, on the grounds that, as banks will have to hold a stock permanently, they need funding for more than the one year stress horizon of the NSFR. This would have destroyed the interbank repo market. Given that the ASF factor of repos for six months or less is 0%, the entire 100% RSF factor would have had to have been covered with capital, making the cost of repo prohibitive. As the same treatment would apply to central bank repo, there is also the question of how monetary policy would be implemented.
Reverse repos with NBFIs
One special target of the NSFR is the asset/liability mismatch created by short-term wholesale borrowing by banks, largely in the repo market, to fund longer-term lending to non-bank financial institutions (NBFIs), in particular, hedge funds. Accordingly, while short-term interbank funding has been assigned an ASF factor of 0%, lending to NBFIs has — since January 2014 — been assigned an RSF factor of 50%, producing a net stable funding requirement of 50% of the transaction.
The problem is that Basel mistakenly sees all SFT as cash-driven trades involving credit intermediation, what is usually referred to as ‘matched-book’ trades. In reality, many SFT are driven, in part or in whole, by collateral. For example, a bank may borrow a security through a reverse repo in order to cover a short position on its own books. This ‘firm short’ could be due to a market-making commitment, a failed delivery from a counterparty or customer, a short risk position or a short hedge. Or, in what could be called a collateral intermediation trade, a bank may borrow the security through a reverse repo from a counterparty or customer in order to lend through a repo to allow a customer who needs to cover his own short position. Although cash moves, this is only as the counterpart of the collateral movement (in effect, the cash is collateralising the collateral).
The difference between matched-book trades on the one hand, and covering firm shorts and collateral intermediation trades on the other hand, is important. Whereas, matched-book trades can create material exposures to liquidity risk, the covering of firm shorts and collateral intermediation trades do not. Thus, in a matched book, there is obvious maturity transformation and consequent liquidity risk in borrowing through short-term repos and lending through longer-term reverse repos. The repos and reverse repos are independent of each other, so there could be a problem in funding the reverse repos, if the repos cannot be rolled over.
In contrast, in covering firm shorts or in collateral intermediation trades, the two transactions are linked. If either of them is unwound, the other can be terminated or not rolled over, so there is little or no possibility of a funding gap.
To the extent there is no mismatch between the two sides of a covered firm short or collateral intermediation trade, there should not be an asymmetry between the ASF and RSF factors.
The impact of the ASF/RSF factor asymmetry on the covering of shorts will make it more expensive, or even prohibitive, for banks and for NBFIs, both leveraged and real money securities investors, to manage their risks. This could have wider consequences for the liquidity of the repo market and therefore for the cost of securities financing and issuance, particularly in the context of other regulatory pressures. The potential impact is illustrated by the scale of banks’ reverse repo business with NBFIs. A special survey by ICMA of banks accounting for about half the business in the regular semi-annual European repo market survey revealed an average share of 37%.
However, the problem for Basel is to find a simple way to carve out the covering of firm shorts and collateral intermediation from the RSF factor for NBFIs. The balance sheet framework of the NSFR may be far too blunt to solve the problems arising from a ‘once size fits all’ policy.
The unnecessary problems caused by the NSFR for linked firm shorts or collateral intermediation trades also affect short positions. The proceeds of short sales (from which a bank receives cash) are given an ASF factor of 0%. Yet, as soon as a short is closed out by buying back the bond, the bank will have a tradable asset. If the asset is a Level 1 HQLA, there should really be no liquidity issue.
The structure of the NSFR will, other things being equal, segment short-term interbank activity into three maturity buckets: up to six months; between six months and one year; and over one year.
As this market segmentation will interrupt bank arbitrage along the curve, will it be reflected in the segmentation of the yield curve, with kinks at six months and one year? And will the shape of curve be affected? If banks are seeking to borrow further out along the curve but lend further in, this would tend to make the curve more concave. However, given the concentration of activity in the very near term, the impact of market segmentation may be delayed.
The simplicity of the maturity buckets also interacts with the ASF and RSF factors to produce some perverse results. For example, lending for less than one year to non-banks is penalised, if it is funded from banks for six months or less: the RSF factor would be 50% and the ASF factor 0%. Perversely, this penalty applies even if the maturities of the borrowing and lending are matched at six months or below, or if a bank borrows interbank for six months and lends to non-bank customers for as short as overnight.
Inconsistencies with the LCR
Under the Liquidity Coverage Ratio (LCR), liquidity risk can be hedged by buying a Level 1 HQLA or funding a reverse repo against Level 1 HQLA, since it is assumed the Level 1 HQLA can be readily liquidated. The NSFR, however, applies an RSF factor of 5-15%. For example, consider a deposit with 100% cash outflow within the 30-day period. An equal value of Level 1 HQLA would cover the liquidity risk. Under the NSFR, on the other hand, if the same deposit was used to buy a Level 1 HQLA in the cash market or through a reverse repo, it would suffer an RSF factor of 5-15% and (given it would have an ASF factor of 0%) a net stable funding requirement of the same amount. If the reverse repo was with a NBFI, there would be a 50% RSF factor and net stable funding requirement.
Alternatively, if the funding from the deposit was lent out as an unsecured overnight interbank, the RSF factor would be reduced to 0% and the net stable funding requirement would disappear. However, the LCR would only recognise 75% of the inflow, because inflows are capped at 75% of the outflows.
While it is straightforward to hedge liquidity risk under the LCR (buy HQLA), the same is not true under the NSFR. Where there is a net stable funding requirement, a bank needs sufficient capital or it will have to borrow surplus cash from a source with a high ASF factor (that means borrowing for more than six months or from non-financials).
Relying on capital is expensive and imposes another leverage ratio. On the other hand, borrowing increases leverage! And it is an additional cost. And that borrowing has to be unsecured, which will attract regulatory capital charges, and needs to be over 30 days in order not to impact the LCR. And investing the borrowed funds in HQLA will lose a haircut in the form of the RSF factor (multiplying up the size of the required borrowing by at least 5-15%).
Will we see new intermediaries enter the market?
It has been suggested that the NSFR will attract new lenders into the repo market. Indeed, the spectre has been raised of a migration to ‘shadow banks’.
One type of candidate might be public sector entities (PSEs). They can offer banks a 50% ASF factor on funding from overnight to one year, compared to interbank funding (0% to six months and 50% only between six month and one year). However, it is difficult to think of PSEs doing this or even of other NBFIs. A more likely prospect is disintermediation of banks through direct interaction between NBFIs and corporates, facilitated by new specialist trading platforms supported by tri-party services. These would be unregulated.
The cost of the NSFR on intermediation
This would be the cost of covering a stable funding requirement with capital or borrowing for more than six months or from non-financials. Assume a bank decides to use 2-year borrowing (for reasons of liquidity management convenience). The calculations are set out in the table below for reverse repos of 7 to 181 days with NBFIs (RSF factor of 505) against German GC funded. The cost of unsecured 2-year senior unsecured debt (assuming the bank is rated A) would be about 3-month EURIBOR + 30bp. This is around 62bp.
The extra cost is 50% of the differential between the cost of funding and the rate of return on the reverse repo. The calculations suggest a bank could have to widen its bid/offer spread to NBFIs by 20-24bp. This would be at least a tripling of the spread.