Stocks have not done measurably better on Fed “POMO” days. Across six Treasury purchase programs from 2009 to 2022, SPY averaged 0.10% on the 1,218 days the New York Fed bought Treasuries and 0.12% on the 432 other sessions of the same programs, a gap far too small to separate from noise. QE2, the program that made POMO a trading-floor word in late 2010, did show purchase days ahead (0.10% vs -0.07%), but on only 27 non-purchase days.
The operation dates come straight from the New York Fed’s public Markets Data API, which lists every Treasury operation from August 25, 2005 (1,909 operations through November 5, 2026), saved in _data/nyfed_tsy_operations.csv. Every number below was calculated from the daily data listed under Sources. More studies are on the Market Stats hub.
What POMO was and why traders watched it in 2010
POMO is short for permanent open market operations. The Federal Reserve Board’s open market operations page describes them as outright purchases or sales of securities for the System Open Market Account, as opposed to temporary repo operations. The New York Fed’s trading desk carries them out.
They became a market topic in 2010. On August 10, 2010 the FOMC said it would reinvest principal payments from its mortgage holdings in longer-term Treasury securities, and on November 3, 2010 it announced a further $600 billion of longer-term Treasury purchases by the end of the second quarter of 2011, about $75 billion a month. The desk published a schedule of upcoming operations, and a popular theory held that dealers who sold bonds to the Fed in the morning put the cash into stocks the same day. The original version of this page, from September 2010, tested that idea on SPY. This version re-runs it on every program since.
The purchase programs in the data
A purchase day here is any session with at least one outright Treasury purchase (bills, coupons, TIPS or floating-rate notes). Agency debt purchases and all sales are left out.
| Program | Window (first to last purchase) | Sessions | Purchase days | % of sessions | Operations | Par bought, $bn |
|---|---|---|---|---|---|---|
| 2005-2007 routine purchases | Aug 25, 2005 to May 03, 2007 | 424 | 58 | 14% | 58 | 66 |
| 2009 Treasury purchases (QE1) | Mar 25, 2009 to Oct 29, 2009 | 153 | 60 | 39% | 60 | 300 |
| Aug-Nov 2010 reinvestment | Aug 17, 2010 to Nov 01, 2010 | 54 | 25 | 46% | 25 | 65 |
| QE2 (Nov 2010-Jun 2011) | Nov 04, 2010 to Jun 30, 2011 | 165 | 138 | 84% | 142 | 778 |
| Maturity extension (2011-2012) | Sep 23, 2011 to Dec 28, 2012 | 318 | 210 | 66% | 219 | 668 |
| QE3 Treasury purchases (2013-2014) | Jan 03, 2013 to Oct 27, 2014 | 458 | 359 | 78% | 364 | 790 |
| Bill purchases (Oct 2019-Mar 2020) | Oct 15, 2019 to Mar 10, 2020 | 101 | 83 | 82% | 95 | 400 |
| Pandemic purchases (2020-2022) | Mar 13, 2020 to Mar 09, 2022 | 502 | 426 | 85% | 587 | 3,178 |
| Bill purchases (Dec 2025-Aug 2026) | Dec 12, 2025 to Aug 11, 2026 | 165 | 52 | 32% | 52 | 338 |
Each window is tied to the FOMC statement that set it up (the QE3 Treasury buying was announced on December 12, 2012 at $45 billion a month, the program was concluded in October 2014, and the January 26, 2022 statement brought net purchases to an end in early March). The maturity extension program bought long Treasuries while selling an equal amount of short ones, so it added no money on net. The data also shows long gaps with no Treasury purchases at all, for example May 3, 2007 to March 25, 2009, and October 2014 to early 2016. During QE2 the Fed bought on 84% of sessions, which leaves very few “other” days to compare with.
Purchase days vs the other days of the same program
Comparing within each program keeps the market backdrop the same: a purchase day in 2021 is compared with a non-purchase day in 2021, not with 2008.
| Program | Purchase days n | Avg | Median | % up | Other days n | Avg | Median | % up | Gap (avg, pts) |
|---|---|---|---|---|---|---|---|---|---|
| 2005-2007 routine purchases | 58 | 0.01% | 0.04% | 53% | 366 | 0.07% | 0.10% | 57% | -0.05 |
| 2009 Treasury purchases (QE1) | 60 | 0.10% | -0.01% | 48% | 93 | 0.26% | 0.37% | 61% | -0.16 |
| Aug-Nov 2010 reinvestment | 25 | 0.27% | 0.19% | 64% | 29 | 0.11% | 0.08% | 55% | +0.17 |
| QE2 (Nov 2010-Jun 2011) | 138 | 0.10% | 0.21% | 59% | 27 | -0.07% | 0.04% | 56% | +0.16 |
| Maturity extension (2011-2012) | 210 | 0.11% | 0.04% | 55% | 108 | 0.02% | 0.08% | 56% | +0.09 |
| QE3 Treasury purchases (2013-2014) | 359 | 0.06% | 0.09% | 57% | 99 | 0.11% | 0.14% | 61% | -0.04 |
| Bill purchases (Oct 2019-Mar 2020) | 83 | -0.03% | 0.16% | 61% | 18 | 0.10% | 0.07% | 56% | -0.14 |
| Pandemic purchases (2020-2022) | 426 | 0.12% | 0.16% | 57% | 76 | 0.18% | 0.15% | 59% | -0.07 |
| Bill purchases (Dec 2025-Aug 2026) | 52 | -0.03% | -0.13% | 40% | 113 | 0.13% | 0.16% | 58% | -0.16 |
| QE-era programs pooled (2009-2014, 2020-2022) | 1218 | 0.10% | 0.09% | 56% | 432 | 0.12% | 0.14% | 59% | -0.02 |
Returns are SPY close to close, dividends included. The three programs of 2010 to 2012 lean toward purchase days, which is the period the POMO idea came from. QE1, QE3 and the pandemic program lean the other way. Pooled across the six QE-era programs (QE1 through the pandemic buying), the gap is -0.02 points and the script’s t-statistic is -0.34, nowhere near the roughly 2 you would want before calling a gap real. The bill programs, which were about managing bank reserves, show nothing either.
Inside the session: open to close
Most operations ran in the morning, so the open-to-close move is the closer test of “the cash goes into stocks the same day.”
| Program | Purchase days n | Avg open-to-close | % up | Other days n | Avg open-to-close | % up |
|---|---|---|---|---|---|---|
| Aug-Nov 2010 reinvestment | 25 | 0.02% | 56% | 29 | 0.05% | 55% |
| QE2 (Nov 2010-Jun 2011) | 138 | 0.04% | 59% | 27 | 0.08% | 52% |
| QE3 Treasury purchases (2013-2014) | 359 | 0.03% | 60% | 99 | 0.02% | 53% |
| Pandemic purchases (2020-2022) | 426 | 0.01% | 55% | 76 | 0.09% | 58% |
Only QE3 shows purchase days ahead on the average, by 0.01 points; in the other three the purchase-day average was the smaller one. QE2 and QE3 purchase days did close above their open a little more often (59% and 60%). During QE2, 133 of the 142 purchase operations opened between 10:14 and 10:16 a.m. Eastern and closed at 11:00 a.m., and three ran from 1:15 to 2:00 p.m. With daily bars there is no way to isolate those windows.
Did bigger purchase days matter during QE2?
If the money mattered, larger operations should line up with larger gains. They don’t line up in a straight line.
| QE2 sessions | n | Avg | Median | % up |
|---|---|---|---|---|
| Under $5bn | 47 | 0.21% | 0.28% | 64% |
| $5bn to $8bn | 75 | -0.02% | 0.05% | 51% |
| $8bn or more | 16 | 0.33% | 0.27% | 81% |
| No purchase | 27 | -0.07% | 0.04% | 56% |
The biggest days did best, but the middle group, which holds most of the purchases, was flat, and the top group has only 16 days. That is not the steady size effect the theory predicts.
What the market did over each program
The table below is context, not cause: it is SPY’s total return from the close before the first purchase to the close of the last one.
| Program | From close of | To close of | SPY total return |
|---|---|---|---|
| 2005-2007 routine purchases | Aug 24, 2005 | May 03, 2007 | 28.11% |
| 2009 Treasury purchases (QE1) | Mar 24, 2009 | Oct 29, 2009 | 33.70% |
| Aug-Nov 2010 reinvestment | Aug 16, 2010 | Nov 01, 2010 | 10.07% |
| QE2 (Nov 2010-Jun 2011) | Nov 03, 2010 | Jun 30, 2011 | 11.63% |
| Maturity extension (2011-2012) | Sep 22, 2011 | Dec 28, 2012 | 27.63% |
| QE3 Treasury purchases (2013-2014) | Jan 02, 2013 | Oct 27, 2014 | 38.92% |
| Bill purchases (Oct 2019-Mar 2020) | Oct 14, 2019 | Mar 10, 2020 | -2.07% |
| Pandemic purchases (2020-2022) | Mar 12, 2020 | Mar 09, 2022 | 77.82% |
| Bill purchases (Dec 2025-Aug 2026) | Dec 11, 2025 | Aug 11, 2026 | 12.74% |
Stocks rose during most programs, which is why the POMO story felt right at the time. But when the Fed bought on 80% or more of sessions, most of any rally will land on purchase days by arithmetic alone. The fair test is the day-by-day comparison above, and there the purchase days look like any other day. If the programs moved stocks, this data gives no sign that it happened on the mornings the desk was in the market; the announcements themselves came on FOMC statement days, which have their own study.
How purchases are reported now
The name has quietly dropped out of use. The New York Fed’s Treasury securities operational details page does not use “POMO” at all: it posts a tentative schedule for each monthly period, announces the day’s operation on its Treasury Securities Operations page when it starts, and posts results after it closes. As of November 8, 2026 that page lists reserve management purchases of Treasury bills from December 12, 2025 (about $40 billion per period at first, then smaller) and none planned for the periods from August 14 to November 13, 2026, with reinvestment purchases continuing. The older POMO page is still up and links to the same operational data. For other recurring calendar days, see our September triple witching study.
All of this describes past sessions. It isn’t a forecast, and none of it is a recommendation to trade around the Fed’s operations.
