SOFR spikes and month-end effects
SOFR is a median of real repo trades, so it moves when the repo market is squeezed. Here are the largest episodes since 2018 and the recurring month-end pattern, measured from the New York Fed's daily data.
SOFR's large moves usually come from Federal Reserve target-range changes, but not always. Its largest one-day move had nothing to do with a policy decision: on September 17, 2019, SOFR printed 5.25%, up 282 bp from 2.43% the day before, and fell 270 bp to 2.55% the next day. At the other extreme, SOFR printed 0.01% on 89 days between March 24, 2020 and June 16, 2021. Between those episodes sits a smaller pattern that recurs every month. Across 97 complete months, the last business day printed on average 5.1 bp above the mean of the five business days before it, and quarter-ends averaged 7.6 bp.
Every figure on this page comes from the Federal Reserve Bank of New York's published SOFR, TGCR and EFFR series (April 2, 2018 through October 8, 2026). Explanations of cause are limited to what the Federal Reserve itself has published, and each one is named where it is used.
September 16–18, 2019: the largest spike on record
SOFR was running at 2.12% to 2.20% in the second week of September 2019, inside a federal funds target range of 2.00%–2.25%. Then it rose two days in a row: to 2.43% on September 16 and, on the second day, to 5.25%, with the whole distribution of trades far above normal:
| Effective date | SOFR | 1st pct | 25th pct | 75th pct | 99th pct | TGCR | EFFR | Volume ($bn) |
|---|---|---|---|---|---|---|---|---|
| Sep 13, 2019 | 2.20% | 2.16% | 2.19% | 2.27% | 2.40% | 2.19% | 2.14% | 1,134 |
| Sep 16, 2019 | 2.43% | 2.38% | 2.42% | 2.55% | 4.60% | 2.42% | 2.25% | 1,156 |
| Sep 17, 2019 | 5.25% | 2.25% | 5.00% | 5.85% | 9.00% | 5.25% | 2.30% | 1,177 |
| Sep 18, 2019 | 2.55% | 2.10% | 2.50% | 3.00% | 5.00% | 2.50% | 2.25% | 1,196 |
| Sep 19, 2019 | 1.95% | 1.82% | 1.92% | 2.05% | 2.50% | 1.90% | 1.90% | 1,137 |
Three details stand out. First, TGCR printed 5.25% too, so this was not a quirk of one segment; tri-party general collateral, the narrowest of the three repo rates, moved by the same amount. Second, EFFR rose only to 2.30%, 5 bp above the top of the target range, so SOFR finished 295 bp above EFFR, the widest gap in the series. Third, volume did not fall. About $1.18 trillion still traded; it simply traded at much higher rates. On September 19 the target range moved down 25 bp to 1.75%–2.00%, and SOFR printed 1.95%.
SOFR against the federal funds target range, 2019
What the Federal Reserve says caused it
The Federal Reserve Board staff note "What Happened in Money Markets in September 2019?" (Anbil, Anderson and Senyuz, February 27, 2020) attributes the pressure to a supply-demand mismatch on September 16. Two things landed on the same day. Quarterly corporate tax payments were due, which moved cash out of bank and money market fund accounts and into the Treasury's account at the Fed. And $54 billion of long-term Treasury debt settled, which increased what dealers needed to finance in repo. According to the note, reserves in the banking system fell by about $120 billion over two business days, to $1.34 trillion, the lowest level since 2012.
The same note discusses why lenders did not step in to take advantage of the high rates. It cites uncertainty and the temporary nature of the spike, and possible frictions from supervisory and regulatory factors.
The New York Fed responded on the morning of September 17. Its Statement Regarding Repurchase Operation announced an overnight repo operation of up to $75 billion against Treasury, agency debt and agency mortgage-backed collateral, to "help maintain the federal funds rate within the target range of 2 to 2-1/4 percent." The FEDS note reports that the operation provided $53 billion of reserves and that rates fell immediately afterward. The published SOFR for September 17 is the volume-weighted median of that whole day's trades, and it still came out at 5.25%.
What one spike did to a compounded rate
A floating-rate instrument that compounds daily SOFR in arrears absorbs a spike like this, but it dilutes it. Compounding plain SOFR from July 1 to October 1, 2019 (92 days, Actual/360) gives 2.2655%. Replacing September 17 alone with the previous day's 2.43% gives 2.2347%. The single print added 3.1 bp to the quarter's rate. On $100 million that is about $7,878 of extra interest over the quarter, almost exactly one day's 282 bp difference at Actual/360 ($7,833). You can reproduce the first figure in the compounded SOFR calculator. The fallout is greater for anything that fixes off a single day's print, and for a short accrual period, because the spike makes up a larger share of the days.
March 2020 to June 2021: the 0.01% floor
SOFR's move toward zero in March 2020 came in steps that tracked two target-range cuts, with an unstable week in between:
- March 4, 2020: the range moved to 1.00%–1.25% and SOFR fell 41 bp to 1.23%.
- March 16: the range moved to 0%–0.25% and SOFR fell 84 bp to 0.26%. That cut followed the FOMC's March 15, 2020 statement, which also said the Fed would increase its holdings of Treasury securities by at least $500 billion and agency mortgage-backed securities by at least $200 billion. The accompanying implementation note set the overnight reverse repo offering rate at 0.00%, effective March 16.
- March 17: SOFR rose 28 bp to 0.54%, above the new range, with a 75th percentile of 0.74%.
- March 18–24: 0.10%, 0.06%, 0.04%, 0.02%, then 0.01% on March 24, the all-time low.
SOFR stayed at 0.01% for 13 consecutive business days, through April 9, 2020. It printed that level on 19 days in 2020 and 70 days in 2021, for 89 in total. The longest run was 68 business days, from March 11 to June 16, 2021. The run ended on June 17, 2021, when SOFR printed 0.05%. That was the effective date of the Fed's June 16, 2021 implementation note, which raised the overnight reverse repo offering rate from 0.00% to 0.05%. The secured vs unsecured rates guide follows that adjustment through EFFR.
SOFR is rounded to the nearest basis point, so 0.01% covers medians from 0.005% up to 0.015%. Trades below zero did happen. The 1st-percentile rate was −0.01% on March 27, 2020 and −0.03% on March 31, 2020. The published rate never went below 0.01%, so a 0% floor on daily SOFR never actually bound in this period, even though some trades were done below zero. Where a loan applies its floor matters only when the published rate itself goes below the floor. The compounding conventions guide covers how daily rates feed an accrual.
The month-end and quarter-end effect
To measure the month-end effect, take the SOFR printed on the last business day of each month and subtract the average of the five business days before it. This site computes that figure for every month from the New York Fed series. Five months are excluded here because a target-range change fell inside the six-day window, so the comparison would measure the Fed's decision rather than month-end pressure: September 2018, October 2019, July 2022, July 2023 and October 2025. October 2026 is excluded because the month is incomplete. That leaves 97 months.
| Period | Month-ends | Months | Mean | Median | Higher than prior 5 | ≥ 5 bp |
|---|---|---|---|---|---|---|
| All months | All | 97 | 5.1 | 2.0 | 78 | 29 |
| All months | Quarter-ends | 33 | 7.6 | 2.4 | 24 | 13 |
| All months | Other months | 64 | 3.8 | 1.9 | 54 | 16 |
| Apr 2018–Feb 2020 | All | 21 | 14.6 | 9.4 | 21 | 15 |
| Apr 2018–Feb 2020 | Quarter-ends | 6 | 26.6 | 22.6 | 6 | 5 |
| Mar 2020–Feb 2022 | All | 24 | 0.5 | 0.0 | 11 | 0 |
| Mar 2022–Sep 2026 | All | 52 | 3.5 | 2.2 | 46 | 14 |
| Mar 2022–Sep 2026 | Quarter-ends | 19 | 4.8 | 3.6 | 16 | 8 |
| Jan 2025–Sep 2026 | All | 20 | 5.4 | 4.7 | 19 | 10 |
| Jan 2025–Sep 2026 | Quarter-ends | 7 | 7.7 | 7.2 | 7 | 5 |
The mean is pulled up by a few large prints, so the median is the better guide to a typical month: about 2 bp. The effect has three distinct phases. Before March 2020 it was large and appeared every month. The biggest jumps were December 31, 2018 (+57.0 bp, SOFR 3.00%), September 30, 2019 (+45.2 bp, 2.35%) and April 30, 2019 (+30.2 bp, 2.76%). During the near-zero period it all but disappeared. Since 2022 it has returned, and from 2025 it has been a regular feature: every one of the last seven quarter-ends printed above its prior five days, led by December 31, 2025 (+15.8 bp, 3.87%) and June 30, 2025 (+10.2 bp, 4.45%). The second-largest month-end jump of the period was not a quarter-end: November 28, 2025 (+14.8 bp, 4.12%).
The jump usually reverses quickly. On the business day after month-end the median change was 0 bp and the mean −2.0 bp. The rate fell in 41 of the 97 months, and big jumps produced big reversals. SOFR fell 47 bp on October 1, 2019, and on January 3, 2019 it fell 45 bp from 3.15%. January 2, 2019 had printed higher than the year-end itself.
Why quarter-ends are different
The Federal Reserve Board staff note "What Happens on Quarter-Ends in the Repo Market" (Bostrom, Bowman, Rose and Xia, June 6, 2025) describes the mechanism. Some dealers "scale back part of their repo intermediation on quarter- or month-ends" to reduce the effect of regulatory or accounting rules on their reported positions, a practice known as window dressing. The authors find that the fall in tri-party borrowing at quarter-end is driven mostly by a few large foreign dealers. They note that European banks calculate their supplementary leverage ratios on quarter-end positions rather than quarterly averages. Some lenders shift cash into the Fed's overnight reverse repo facility on those dates. The note adds that repo rates tend to rise on Treasury bill or coupon issuance days, and that the effect can be larger when those days fall on a quarter-end.
From September 2025 through year-end, SOFR printed above the top of the target range on 17 days, against 2 days in all of 2024. The widest SOFR–EFFR gap of 2025 was 36 bp on October 31, 2025 (4.22% against 3.86%). The data show when it happened. They do not explain it, and this page does not offer a cause for it. The daily prints are on the SOFR in 2025 page.
The largest daily moves since 2018
Ranking every day-over-day change since April 2018 by size mixes two kinds of move. Nine of the fifteen largest happened on a day the federal funds target range changed, and those are policy moves passing through. The other six, marked "no", came from funding pressure on the dates shown.
| Effective date | From | To | Change | Target range changed that day? |
|---|---|---|---|---|
| Sep 17, 2019 | 2.43% | 5.25% | +282 bp | No (mid-September repo pressure) |
| Sep 18, 2019 | 5.25% | 2.55% | −270 bp | No (reversal) |
| Mar 16, 2020 | 1.10% | 0.26% | −84 bp | Yes, to 0%–0.25% |
| Jun 16, 2022 | 0.70% | 1.45% | +75 bp | Yes, to 1.50%–1.75% |
| Jul 28, 2022 | 1.53% | 2.28% | +75 bp | Yes, to 2.25%–2.50% |
| Nov 3, 2022 | 3.05% | 3.80% | +75 bp | Yes, to 3.75%–4.00% |
| Sep 22, 2022 | 2.25% | 2.99% | +74 bp | Yes, to 3.00%–3.25% |
| Sep 19, 2019 | 2.55% | 1.95% | −60 bp | Yes, to 1.75%–2.00% |
| Dec 31, 2018 | 2.46% | 3.00% | +54 bp | No (year-end) |
| Sep 30, 2019 | 1.82% | 2.35% | +53 bp | No (quarter-end) |
| Dec 15, 2022 | 3.80% | 4.32% | +52 bp | Yes, to 4.25%–4.50% |
| Sep 19, 2024 | 5.33% | 4.82% | −51 bp | Yes, to 4.75%–5.00% |
| May 5, 2022 | 0.30% | 0.79% | +49 bp | Yes, to 0.75%–1.00% |
| Oct 1, 2019 | 2.35% | 1.88% | −47 bp | No (quarter-end reversal) |
| Jan 3, 2019 | 3.15% | 2.70% | −45 bp | No (year-end reversal) |
Target-range dates here are the first EFFR publication in each new range, as listed on the target range history page. Every move in the table that was not a policy move falls between December 2018 and October 2019. Outside target-range change days, the largest move since then was March 18, 2020 (−44 bp, to 0.10%). After that date, no move without a target-range change has reached 20 bp. The largest was +18 bp on October 31, 2025 (4.04% to 4.22%), followed by +16 bp on December 31, 2025.
Checking a period for spikes before you rely on it
When a lender's compounded rate differs from your own, or a monthly average looks off, look first at the month-end and quarter-end prints in the period. The SOFR history lookup lists every print in a date range. The percentiles and volume page shows whether a high print came with a wide 99th-percentile tail, as December 31, 2018 did at 6.25%. For a period from March 2, 2020 onward that spans a known spike, the SOFR Averages and SOFR Index already include it, so a calculation from the Index should match one compounded from the daily rates. The New York Fed began publishing both on March 2, 2020, so for the 2018 and 2019 spikes there is no Index value to check against; compound the daily rates in the compounded SOFR calculator instead.