Is SOFR better than LIBOR?

In this blog post, I summarise the mechanics and whys behind the choice of discounting rates, collateral management and evaluate whether the move to SOFR is an improvement over LIBOR. In the next article, I will write on what banks can do to transit from fixed rate, fixed term loans and deposits to floating rate, fixed term ones i.e. restructuring of bank’s balance sheet with credit derivatives.

Collateral rate

Assuming ideal “perfect collateralisation” conditions, the standard collateral rate specified in a collateral agreement is the OIS rate.

What makes deals covered by “perfect” collateral agreements?

  • 2-way (if its 1-way, favored party receiving collateral has to pay the collateral rate, whereas the disfavored uncollateralised party will receive funding benefits but will be unprotected from any jump risk)
  • symmetric MTM (e.g. for an option deal, payoff is asymmetric and the seller (ITM) of the option that expires OTM do not have any MTM exposure from the buyer. Vice versa, if option is early exercised or expires ITM, the seller (OTM) has to settle the replacement cost it owes to the buyer)
  • 0 threshold and no MTA (possible if only cash is exchanged as collateral)
  • collateral posted/received real-time (possible if collateral can be deducted from a segregated amount, not for default/jump risk i.e. initial margin. Would it make sense for both sides to overcollateralised the other? Typically, the CCP or deal makers providing access have the bargaining power. E.g. CCP initial margin from clients, CCP default funds and intraday collateral posting)
  • no close-out costs besides replacement costs (MTM)
  • collateral is settled in cash in the currency of the deal (if 2 way collateralisation is in different currency, collateral rate could be based on mismatched OIS rates. Resulting mismatch e.g. FF-EONIA could be closed with CCBS spreads.

Why OIS rate? OIS rates are an average of deals that actually occurred and are published timely in the evening (Euro) or the next day (US). OIS rate is a rate that is set rather than a rate available for investment. OIS daily timely fixing is aligned with daily collateral posting, hence OIS discounting is the appropriate discounting rate, given “perfect collateralisation” conditions.

OIS formula
Formula for n days OIS rate

Note: Compounding daily means interest accrued is re-invested daily over floating period. It is equivalent to if interest were paid daily, by no arbitrage argument.

OIS curve
OIS curve is a floating curve, compounded ON to a a floating term rate. LIBOR is a flat term structure fixed before contract.

Compared to IBOR – a banks’ poll on a certain day what they “estimate” uncollateralised lending rates (unsecured > secured rate) to be that they are willing to lend based on their currency liquidity supply, or from a demand perspective, LIBOR reflects “offshore” deposit rates (eurodollar, euroyen,…) that each bidding banks believe it can attract lenders and borrow funds in reasonable size should they seek to do so (typically, since offshore allows banks to avoid regulation applicable to onshore deposit markets, offshore > onshore deposit rate) – OIS rate derived from real deals is more risk-neutral accounting for all participants involved in the OIS market, whereas IBOR is untimely where submission could be biased since panel banks are incentivise to manage excess liquidity or meet near term funding needs. IBOR fixing date typically starts 2 business days forward of the start of rate accrual period, resulting in fixings slower to adjust to rate changes than real-time derivatives markets and wider bid-ask spreads as confidence intervals that likely remains valid the next day, or if invalid, fixing panel banks can adjust rates in small increments. This fixing delay structure means on some weekdays (Wed), volatility in fixing changes could be greater than other days.

That said, from the calculation methodology of InterContentinal Exchange LIBOR, each panel bank are moving away from basing their submission from the typical LIBOR Submission Question “At what rate could you borrow funds, were you to do so by asking for and then accepting interbank offers in a reasonable market size just prior to 11 am?” towards basing submission on the Waterfall Methodology that’s more risk-neutral based on concrete real deals, and less subjective. Note: banks typically submit rates only for non-domestic currencies, unless they are a foreign branch. There’s 5 currencies, each spans 7 tenor points.

LIBOR submissions under the Waterfall Methodology

Why is collateral rate the discounting rate?

A positive MTM will result in collateral received, and this collateral balance will grow according to the collateral rate paid by the collateral receiver. Final collateral balance (variation margin) should be able to replace MTM exposure if collateral giver defaults. In other words, collateral is the amount which, if re-invested today, would cover all expected payments in the future. So by symmetry, collateral rate is the correct discounting rate on MTM. Since any FV will be a replacement cost before deal is closed, it goes to say collateral (OIS) rate is the correct discounting rate.

collateral rate
Collateral rate = MTM discounting rate

Its simple when PV is projected with the same rate to FV as when FV is discounted to PV. However, deposits/loans reference rate are typically quoted in LIBOR as they captures the deposit/loan maturity term premium and the counterparty risk they are exposed to when they deposit money in banks (i.e. banks only provides deposit insurance guarantees in case they default, which insurers are typically the government) or loan money to corporate/institutional clients. One would ask, why not use any cash collateral posted to purchase fixed deposits with LIBOR as reference rate? For many reasons from credibility to A-L maturity mismatch, the collateral receiver cannot do that and has to seek the most liquid investment which is the OIS referencing the overnight rate. That said, projection from PV to FV is done with LIBOR as the minimum expected risk-free return, while discounting from FV to PV is done with OIS as it’s the popular collateral agreement referenced rate.

However, like collaterisation in different currencies, the use of 2 rates curves would lead to asymmetry in MTM and collateral balance growth. LIBOR-OIS basis swap can be used to “basis adjust” the OIS floating leg to capture the risk and term premium that LIBOR compensates.

That said, LIBOR (USD) as reference rate to many fixed deposits/loans would be largely replaced by or complement with ON risk-free rates. The trend to move away from IBOR to ON risk-free rates is seen globally. In the US, the Alternative Reference Rates Committee (ARRC) has recommended the Secured Overnight Financing Rate (SOFR) as the LIBOR replacement for derivatives. SOFR is the combination of three overnight treasury repo rates. According to the Loan Syndications and Trading Association (LSTA), SOFR may become the replacement rate for cash products, like syndicated loans and CLOs. In the UK, Reformed SONIA (Sterling Overnight Index Average) has been identified as the appropriate replacement for GBP LIBOR. In Switzerland, SARON (Swiss Average Overnight Rate) has replaced the TOIS benchmark. In Japan, TONAR (Tokyo Overnight Average Rate) has been selected as the alternative to yen LIBOR. Finally, the ECB has decided that its newly developed unsecured overnight rate would be called ESTER (Euro Short-Term Rate). Note that each country’s central bank has its own treatment of its policy rate.

This would address the issue of IBOR with untimely fixing to capture the term premium accurately, its subjective submission from panel banks and the lack of liquidity in interbank lending rendering IBOR unstable.

Note that central banks play an active role in setting ON rate targets and as such, there is certainty that ON rates will be stable (of course, only if central banks are credible with their promise to meet their targets) and these dates are predictable (OMO dates).

Would this be a perfect solution? Not really. One might question if banks are willing to benchmark loan and deposit rates on repo rates set by the repurchase market.

What is repurchase trades?

Cash borrower can post $A collateral to borrow $B amount (B<A), where collateral’s market value is discounted by a haircut (provides collateral buyer an initial margin). At some future date, the cash borrower (collateral seller) must repurchase the collateral at a predetermined amount $C. The actual repo rate or financing cost is reflected in the future invoice price. If the collateral has a futures market and a futures price that coincide with the repurchase date, there could be an arbitrage opportunity when the actual repo rate of the Cheapest-To-Deliver (cash) bond collateral is lower than the repo rate implied by the futures bond price. This can also be seen when the net basis (basis after carry) is positive, reflecting that futures price is overvalued (pricing a carry at a higher implied repo rate) relative to carry price (at actual repo rate).

How to arb? Carry (buy collateral, repo in, lend cash) the cash bond (at dirty bond price) and finance bond purchase at actual repo rate lower than the repo rate implied by the futures bond price (repo out, seller has to repurchase) i.e. reverse-repo agreement collateral repurchase price should be similar to the futures bond price now.

actual vs implied repo.png

If the collateral is an equity, the repo is called an equity repo. Just as accounting accrued interest between coupon payments (dirty price) and conversion factor compensates for the difference between coupon payment and and bonds delivery, corporate actions does affect equity repo trades. Rights issue will dilute collateral value and the collateral seller is called to make a decision, dividends are paid net of withholding tax, equities are more volatile and less liquid than government bonds, resulting in higher equity repo rates. Idea of GC can be applicable to the country’s top member equity index. Typically, collateral made up as a basket of stocks and if stock falls out of the index basket, there’s a substitution to replace with the new joiner or another top member security. And if market value of collateral falls significantly such that variation risk exceeds threshold, there will be a call for collateral posting or top up of variation margin.

How did repo came about?

The incentive to lend out cash without creating any credit risk for itself. Repo was the ‘brillant’ invention of Benjamin Strong, then President of the Federal Reserve Bank of New York. As a tool of monetary policy, the fed can lend to commercial banks and increase their Fed liabilities in exchange for 100% Treasury collateral. A Treasury primary dealer might object to such a steep collateral requirement, but there is only one source of Fed money, so they repo out Treasuries to get central bank funding. However, when the central bank needs funding or intends to drain liquidity from the market, they carry out reverse repo where the Fed provides 100% collateral while receiving cash. Retail loan takers are deemed more risky than banks and hence borrowing money from banks are always at higher rates than if you deposit fix with banks. Does it makes sense for the Fed to post collateral, when the Fed can creates cash so it can never default on its own bonds? Actually, post WWI, cash was backed by the gold standard and a surplus of gold led to a surplus of cash, resulting in inflation. Strong intends to counteract the surplus of gold and break the relationship between the the Treasury’s stock of gold and the quantity of money circulating onshore.

Incidents involving the misuse of repo that shaped repo agreements now:

  • [1972] In response to the 1968 Wall Street Paperwork Crunch where the paper-based system left plenty of room for mix-ups and foul play, resulting in missed trades, brokerage failures, market shutdowns, and theft, the SEC enact Rule 15c3-3, designed to protect client accounts at securities brokerage firms. In short, the rule dictates the amount of cash and securities that broker-dealer firms must segregate in specially-protected accounts on behalf of their clients. The cash and securities segregated therein cannot be rehypothecated by the firm for any purpose, such as trading for its own account or funding its operations. The intent is to ensure that clients can withdraw the bulk of their collateral posted on demand, even if a firm becomes insolvent. This led to breakthroughs with “central stock certificate system” and what we call Central Securities Depositories (CSD).Currently, broker-dealers in the U.S. cannot use equities as collateral in Securities Financing Transactions (SFT). Nowadays, securities borrowing are financed by half cash and half non-cash collateral. Given that most equity collateral driven businesses are based outside the US, with the highest concentrations being in Europe, Asia Pacific and Canada, a change in Rule 15c3-3 will likely result in a shift back towards the US. Furthermore, under Basel III, there is a 100% capital charge on agent lenders when they lend equities to non-bank entities that in turn deliver equities as collateral (rotation of equities flow).
  • [1982] Drysdale Government Securities exploited the differences in accrual interest treatment between ON repo where AI seemed insignificant overnight and securities lending where AI was accounted – Drysdale used the Treasuries it had reversed repo in (at clean price) to make outright short sales to a third party for an amount that included the accrued interest. Using the surplus cash generated, Drysdale was able to raise working capital and to make interest payments to its other repo counterparties. Overall, the strategy wins under two 2/3 possible market scenarios. If the bond market went down, AI amplifies. If bond market stayed the same, he earned free interest on the cash the trade generated from AI arb. If the Treasury market rallied, he risked a significant amount of money. As it turned out, between February 1982 and May 1982, the long-end of the Treasury market rallied considerably. Cumulative losses on Drysdale’s interest rate bets caused it to be unable to pay the interest on securities it had borrowed. Lesson learnt is that despite ON, repo has to include AI and “full accrual pricing” became standard market practise.
  • [1982] Lombard-Wall had Repo transactions with the Street and with customers, but in this case, there was no large bank to cover the losses. Immediately following the bankruptcy, Lombard-Wall’s Repo trades were tied up in bankruptcy court with no one allowed to liquidate them. In September 1982, one month later, the Federal Bankruptcy Court of New York ruled that a Repo was a buy and a sell, making it two separate transactions and thus not a collateralized loan. The court recognized that allowing prompt liquidation was necessary to continue the orderly functioning of the markets. Two years later, in 1984, Congress passed an extension of the Federal Bankruptcy laws so that Repo on Treasuries, Federal Agencies, CDs and BAs were exempt from automatic stays in a bankruptcy by law.
  • [2008] Despite running out of collateral to back short-term refinancing,
    Lehman managed to survived by window dressing its balance sheet with Repo 105 (assets moved amounted to 105% or more of the cash it received in return) where the “sale” (rather than financing) is merely collateral posting that have to be repurchased later. These Repo 105 transactions often were done in flurries in a financial quarter’s waning days, before Lehman reported earnings. At the same time, Lehman was able to double-count some liquid securities by pledging them as collateral in repo agreements and counting them in its internal liquidity pool (that can be monetized at short notice in all market environments).
Complex Repo 105 illustration
Detailed illustration of a Repo 105 transaction

Tri-party repo

With CSD to keep a clear, centralised record of the multitude of paper contract ownership comes clearing and settlement fees as extra overheads to transact across CSD for collateral pledged in repo agreements, and this refinancing happens frequently magnifying the overhead cost. As such, (Hold-In-Custody) HIS repo,where possession of collateral (although not ownership) remains with the collateral seller (pledger), are favoured because its cheaper than delivery repo and makes it economic to repo out less liquid collateral which has lower turnover and way more expensive fixed settlement fees. However, the downside to HIC repo is that the buyer (collateral taker who lend out cash) will not be able to monitor further movements to the collateral pledged or would the pledged collateral be unsegregated and repo out to more than one seller. This led to a similar concept of centralised depository for collateral pledged in repo agreements, whereby they system is tailored for handling the more complex and frequent change of ownership of collateral – a tri-party agent.

Image result for triparty repo agent
Tri-Party Transaction Flow with agent BNY Mellon (who has about 80% market share of triparty repo in the U.S. domestic repo market)

In a tri-party repo, the securities of both buyers and sellers are held by the agent in an ‘omnibus‘ account (joint account with possibly more than 2 counterparties to simplify pledging and receiving operations) at a CSD, which can be domestic or international e.g. DTC, ClearStream, Euroclear. When a transfer of securities ownership takes place, there’s no movement in securities across CSD but a simple change of ownership record by book-entry. This arrangement makes the cost as cheap as a HIC repo but also ensures collateral segregation. As mentioned, securities that have high fixed settlement cost due to illiquidity is equally favored in tri-party repo as in HIC repo. Beyond cost effectiveness, tri-party agent goes beyond to enforce better collateral management such as facilitating DvP that reduce settlement risk or delayed settlement than a CSD for later refinancing.

In the US, both side shares the fee paid to agents, while in Europe, agent fee is paid only by the seller (typically securities dealers) who seeks to attract an unfounded source of funding from the central bank and repo market that would otherwise not be made accessible because of their lack of operation requirements for collateral management and securities settlement. Tri-party is therefore the easiest route to access repo market for cash investors beyond the dealers community (e.g. insurance companies, money market funds, sovereign wealth funds, large corporate treasuries, commercial banks, pension funds, securities lending agents with excess cash collateral to reinvest). More innovative products in the U.S. led to collateral types such as agencies (and most “credit repo” involving corporate bonds, equity, covered bonds, structured securities, ABS, MBS, CDS where operations lack familiarity) that are more complex to manage, resulting in the delegation of their management to agents. The sheer size of the Agency repo market in the U.S. also meant that the agent fee is cheaper, resulting in delegation of other products too such as US Treasury. Triparty repo makes up 60% of the repo market in the U.S. compared to 10% in the Europe.

Don’t be mislead by the word “tri-party” as the agent are not principle intermediaries in the collateral transactions they manage. Agents are not exposed to any counterparty risk as they have no claims on collateral ownership but whose sole responsibly is manage collateral. In the event that either one counterparty defaults, it simply refuse further instructions from the defaulter, segregate existing collateral and stop further collateral exchange, while waiting for non-defaulter instructions.

Main function of a tri-party agent:

  • At start of transaction: select securities to be delivered as collateral from pledger to buyer account.
    Collateral selection in tri-party repo can be manually done by pledger or automatically by an algorithm:

    • Eligible as collateral by buyer – sufficient quantity in one or more securities available to cover credit and liquidity risk of collateral, fits within buyer’s risk appetite and concentration limits on their balance sheet
    • “Cheapest-To-Deliver” (CTD) collateral from the pledger’s collateral pool managed by the agency. CTD to seller are typically collateral of lowest credit and liquidity quality or those held in inconvenient amounts, as seller would like to improve their liquidity and credit risk profile. That said, buyer can demand to receive a higher repo rate if the collateral pledged are of lower quality. CTD collateral can be a basket (~40 pieces of collateral) and since their specific identity is irrelevant, such cash-driven repo is known as GCs
  • During life of transaction: select securities to deliver as be margin or substitutes from the account of one party to another. Typically, the pledger has unlimited rights of substitution of collateral.
    Collateral are substituted by the pledger whenever:

    • collateral pledged have been sold by the pledger and has to be retrieved for delivery in settlement of that transaction
    • delivered collateral became ineligible e.g. ratings downgrade to a rating below the eligible threshold set by buyer
    • (Optimisation) pledger receives a new security that is cheaper to deliver than the delivered collateral
    • income is due to be paid on the to-be delivered collateral i.e. e.g. Collateral buyer can avoid an income payment to avoid triggering tax consequences by substituting before record date.
    • (Fail-driven optimisation) delivered collateral are required for delivery to another tri-party repo to avoid a delivery failure
  • Issue settlement instructions to CSD:
    • At start of transaction: DvP (Delivery vs Payment) of selected securities to the buyer’s and seller’s tri-party collateral account
    • At end of transaction: Delivery Free-Of-Payment in cases where the collateral have been substituted or margined
  • Monitoring collateral value against cash owed to buyer
    • when collateral value falls short of the cash owed to the (collateral) buyer, pledger have to top up collateral by transferring margin to buyer; vice versa, when collateral value exceeds, buyer returns excess collateral by transferring margin to pledger
    • If collateral is net crediting, income collected by (collateral) buyer have to make a corresponding “manufactured payment” to the pledger.

Note: In the US, agents are not involved in collateral selection, substitution, margining or income payment. There functions are replaced by a daily cycle in which repos are automatically unwound and recontituted. This daily cycle provides opportunities to pledgers to

  • withdraw rather than substitute later
  • collect any income payments directly rather than wait for for manufactured payments from buyers
  • reassemble an entirely new portfolio of collateral that is cheapest when valued at the latest prices rather than margining

Collateral buyers just concern themselves with whether the new portfolio collateral value is sufficient to cover the cash owed to them.

In this sense, the role of agents in US is largely about cost-efficient custody of frequently delivered collaterals and DvP settlement. However, the US triparty repo system is moving towards that in Europe as there is a large settlement risk in the risky interval between unwinding and reconstitution.

Challenge of segregation in tri-party books

An operational challenge is collateral rehypothecation (segregated = “N”). If the buyer reuses the collateral as a pledger to another buyer (e.g. a new buyer has a need for the collateral and are willing to accept a lower repo rate, thereby lowering financing cost), therein recovering the cash it loan out as a collateral buyer but earning from the difference in repo rate as a collateral seller in a reverse-repo, how can the triparty agent be sure that it can retrieve the collateral when the seller wants to substitute the security that have been repo out by the buyer? If there are legals stating that if the buyer cannot reuse the collateral, full legal title have not been transferred to the buyer, in which this undermines the buyer’s confidence in retaining the collateral in the event of default by the seller, triparty agent will circumvent this legal issue by allowing buyers to reuse, but if doing so, transaction would ceased to be a triparty repo and will no longer be maintained by the agent. This tri-party segregation of collateral became an regulatory problem as regulators are unwillingly to allow triparty repos to be included in the liquidity ratios of the buyers i.e. If i repo in a high-quality collateral that are typically as liquid as cash since i can repo out with another buyer, they can be confidently said to be liquid to be included by the regulators.

This issue is resolved with the reuse of facilities whereby the same triparty agent is used when the buyer post someone’s collateral as collateral in another tri-party repo. In this way, the agent will know where to find the collateral when any one of the pledger decides to substitute it.

So the pledger can either substitute it and directly reverse-repo it out to the buyer who needs the collateral and lower financing cost, or allow the buyer to reuse it. Is collateral did became special and highly demanded, the former (substitution) favors the pledger as they get to keep the lower financing cost. As much as pledgers can optimise their portfolio of collateral posted, the bargaining power lies with the buyer who has the cash to offer short term cash financing. If the collateral buyer insist on more liquid collateral substitutes, pledger have to conform to its minimum collateral eligibility or if not, be unable to get financing.

Transition from IBOR to ON repo (aka collateral) rates

ISDA renegotiation to use OIS as reference rate over LIBOR and the banks willingness to make a market on ON repo rate futures based on some benchmarking or indexing make a statement of support towards using ON repo rates as the reference discounting rate in all derivative, loan and deposits contracts. However, ON repo rates will dilute the interbank position as the main intermediary of cash transmission who used to determine the money velocity in the economy, as rates would be set not on the interbank transactions-based data, showing demand and supply of cash through the banking channels, but on the repurchase market where there is more cash investors besides the banks and the central bank liabilities who are willing to provide short term financing by securitising their assets. There would be more financiers able to yield the credit risk premium in the business of lending and borrowing (cash rich investors need not borrow and lending becomes an investment). However, like in the buying and selling of Treasuries as a popular collateral, there eventually need to be a rational way to guarantee that the fair value of the collateral is actually real and acceptable to the majority, thereby affirming its fair value. Otherwise, it would become another re-enact of 2008 MBS GFC. That said, there are appropriate credit risk policies such as KYC and credit approvals before loans are made, deciding on eligibility of collateral in repurchase agreements, using risk-neutral market prices in measurement models to reflect everyone’s opinion, conservative haircuts based on history and experience on market value to factor that market could be entirely wrong too. This in turn would protect their own creditworthiness and those exposed to them too.

Given that credit risk have largely changed from a fixed term of a few months to overnight or days, there’s a view that credit risk have largely transformed to operational risk, whereby good credit risk monitoring and collateral management practice that is more reactive (and predictive). No more subjective submission of benchmark rates by the few monopolistic panel banks. A more engaging way of determining the benchmark whereby more financiers are allowed to participate in the demand and supply of cash by taking risk to make a democratic statement of where rates should be. That said, operational pressure on a few centralised agencies who provide access to these non-bank financiers and their enforcement of collateral posting and retrieval would grow as the move to a “perfect collateralisation” environment underpinning the setting of a representative benchmark rate increases the frequency of collateral management.

Would the next “Paperwork Crunch” as a result of operational defects or failures be justifiable for a more “representative” benchmark rate over a multi decade old subjective submitted rate? In the end, there still needs to have a “fixing” leg and someone to take the risk premium. LIBOR has to stay as a complement as banks have the edge to hedge the term risk by restructuring their assets and liabilities with basis swaps. Basis swap spreads will also grow as an indicative telltale sign of the hedging cost and hence the willingness for financiers to engage in the repurchase market.


  • The ICMA Guide to Best Practice in the European Repo Market
  • ICMDA Repo FAQ

One Reply to “Is SOFR better than LIBOR?”

  1. How did you write an entire piece on SOFR without referencing Basel III’s LCR and HQLA requirements??? Banks do not borrow unsecured anymore, and haven’t for some time, this is NOT a voluntary transition or a matter of opinion, there is virtually no volume in 3m Libor, and the Fed and FCA have made this neccessity abundantly clear.

    Libor will not exist after 2021, there may even be legal measures which force this, but either way it’s not a funding rate. NO BANKS BORROW AT IT.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s