RFR lookback, lockout and observation shift conventions are the rules that decide which days' rates are used, and how each is weighted, when interest on an overnight risk-free reference rate is calculated. Rates such as SOFR, SONIA and the euro short-term rate are published each day for the previous day, so a rate compounded in arrears over an interest period would not be known until the period ends. Borrowers, agents and paying agents need the amount a few days earlier to arrange payment.
The conventions solve that problem by moving the window of rates backward or by freezing the final rates. They were standardised during the LIBOR transition by industry working groups, the Loan Market Association and ISDA. The choice of convention is a commercial term, and the differences between the methods are usually small but not zero. The arithmetic is set out more fully under RFR compounding methods.
Observation Period
The observation period is the run of days whose RFR values are compounded to give the rate for an interest period. Under a plain in-arrears method the observation period and the interest period are the same, which leaves no time to calculate the payment. Under a lookback or observation shift, the observation period starts and ends a set number of business days before the interest period does: a five-day lookback on an interest period from 1 March to 1 June uses rates from roughly five business days before 1 March to five business days before 1 June.
The observation period still contains the same number of rates as the interest period would, so the borrower pays for the full term, just measured slightly earlier. Its exact dates depend on the business day convention and the holiday calendar of the RFR's currency, which the contract must specify.
Lookback
A lookback is the convention of taking each day's RFR from a fixed number of business days earlier, so that the compounded rate is known shortly before the payment date. It is needed because overnight RFRs are backward-looking: unlike LIBOR or Term SOFR, they do not give a rate at the start of the period. With a five-business-day lookback, the last rate used is the one for five business days before the period ends, giving the agent time to calculate and notify the amount due.
Five business days became the common choice in the loan market, reflecting operational timetables in syndicated lending, but it is negotiable. Bonds and derivatives often use shorter gaps, and some loan deals use three or two days. The term "lookback" on its own does not say how the rates are weighted; that depends on whether an observation shift applies, so loan documents specify both elements.
Lookback Without Observation Shift
Lookback without observation shift, often called the lag method, takes each day's rate from the shifted earlier date but weights it by the number of calendar days that the corresponding day in the interest period covers. If a Friday in the interest period is followed by a weekend, the rate observed for it receives a weight of three days, even when a holiday in the observation window means the observed day was a midweek day that applied for only one day.
This keeps the day count tied to the interest period, which suits syndicated loans where borrowers prepay, draw and repay on any day and lenders trade participations. It is the common loan-market method for SONIA and was widely adopted for SOFR loans. Its drawback is a small mismatch with the observation shift used in many hedges, because the same rates can be weighted differently. The mismatch is generally a few basis points or less over a period and is often accepted for operational reasons.
Observation Shift
An observation shift, used in a lookback with observation shift, moves both the rates and their day weights to the shifted observation period. Each rate is weighted by the number of calendar days it actually applied in the observation period, not in the interest period. The compounded result is therefore the true compounded rate over a slightly earlier window, and it can be calculated from a published compounded index for that window.
This method is common in floating rate notes and in derivatives. The ISDA fallbacks for interest rate swaps that referenced LIBOR, for example, compound the RFR with a two-banking-day backward shift. Because it matches published indices and averages, it is simpler to reconcile and audit. Its disadvantage in loans is that the weights can diverge from the days in the interest period around holidays and month-ends, which complicates intra-period prepayments and transfers.
Lockout
A lockout freezes the daily RFR for the last few business days of an interest period at the value observed on the lockout day, so the full amount is known before payment. If an interest period has a two-day lockout, the rate for the final two days is set equal to the rate on the day before the lockout starts, while earlier days use their own published rates without any backward shift.
Lockouts were used in some of the first SOFR floating rate notes because they keep the observation period aligned with the interest period. They are less suitable for loans, because a borrower that repays inside the lockout window has no clean way to settle interest, and the frozen days create a small exposure to rate moves that hedges do not share. The ARRC conventions for SOFR loans therefore favoured lookbacks.
Compounding formula and worked example
Whichever convention applies, the compounded rate for an interest period is calculated as:
- is the RFR for business day , taken from the date the convention specifies.
- is the number of calendar days the rate is weighted for, counted in the interest period (lag) or the observation period (observation shift).
- is the number of business days, the number of calendar days in the period, and the day count base, 360 for SOFR and €STR and 365 for SONIA.
Take a four-day period on a 360-day basis with a Friday rate of 5.00 per cent weighted for three days and a Monday rate of 5.10 per cent weighted for one day. The product is , so per cent, slightly above the simple weighted average of 5.0250 per cent. If an observation shift had moved the three-day weight onto a different rate, the answer would change, which is the source of the mismatch between the methods.
Ref 1100001635