# Overnight reverse repurchases and the Fed's balance sheet ![Screenshot 2025-09-07 at 2.50.58 PM](https://hackmd.io/_uploads/HJgVELjqxe.png) $\cdot$ **tldr;** This article takes a high-level look at the dominant Fed balance sheet trends over the past 2.5 years (since March 2023). We start simply by exploring the Fed liabilities and observe that, while today's liabilities are dominated by the trifecta of reserve balances, physical currency, and the Treasury general account, these three components alone do not account for the *2 trillion dollar reduction in the liabilities* that has taken place over the short term. Nearly the entire drawdown of the liabilities is due to the gradual emptying of the [Overnight Reverse Repurchase Agreement](https://fred.stlouisfed.org/series/RRPONTSYD) facility (shown over this time span in the image above). The decrease in liabilities is actually driven by the runoff of treasuries and mortgage-backed securities (quantitative tightening) happening on the *asset side of the balance sheet*. We demonstrate the mechanics of these capital flows. To conclude, we highlight some of the Fed's own writing on this topic. $\cdot$ *by [mike](https://x.com/mikeneuder) on September 9, 2025.* $\cdot$ <sub>*It is the $252^{nd}$ day of 2025, which is super funny because this [tweet](https://x.com/mikeneuder/status/1955667625855639957) was from the $225^{th}$ day of 2025. I swear none of this was on purpose; just a good time for 2s and 5s.*</sub> --- > **Disclaimer** I know, I know... the last thing you want is another crypto person talking about the Fed and monetary policy – I get it ... But I am going to write about it anyway, for a few reasons. > 1. *Personal learning:* I find that writing, and in particular the process of linearizing a bunch of disparate pieces of data into a cogent narrative, forces me to understand the details of a topic in a way that I wouldn't otherwise. > 2. *Maybe someone will care:* A lot of content about the Fed and their policies is veiled in jargon (maybe intentionally). I hope to present some of these topics in a very ground-up fashion that will probably be more accessible by nature of my not knowing very many macro things. > > But of course, read everything below with a large grain of salt. I am sure there are inaccuracies and oversimplifications, as I am only just beginning to dip my toes into understanding how the Fed operates. YMMV. --- ## Fed Liabilities We start on the liabilities side of the balance sheet because my working definition for a paper dollar[^1] is simply, "a liability on the Fed's balance sheet." The figure below shows the Fed's liabilities since 2003. [^1]: Of course, paper dollars are only a small part of the "dollars" in existence. In particular, paper money is part of the "monetary base" (M0), and is almost an order of magnitude smaller than the amount of dollars held in fractionally-reserved checking accounts (M1)[^2]. Still, for paper money, thinking about them as purely Fed liabilities is useful IMO and you have to start somewhere ¯\_(ツ)_/¯ [^2]: You wanna see a crazy COVID chart? Try this on for size ([source](https://fred.stlouisfed.org/series/M1SL)). ![Screenshot 2025-09-07 at 3.46.46 PM](https://hackmd.io/_uploads/B1WHZPo5lg.png =400x) ![liabilities](https://hackmd.io/_uploads/HyxtXwocge.png =700x) (see the FRED [WLTLECL](https://fred.stlouisfed.org/series/WLTLECL) series.) What drew my attention was the steep decline starting in mid-2022 and resuming after the brief [2023 banking crisis](https://en.wikipedia.org/wiki/2023_United_States_banking_crisis). I knew the Fed's balance sheet was shrinking, but I wanted to understand the details further. As such, we will focus on the time series starting March 1, 2023, shown below. ![liabilities](https://hackmd.io/_uploads/Byq04vs9xx.png) Let's unpack this relatively fast 2 trillion dollar ($\approx$ 24%) decline. Starting with today's numbers, we can cleanly break down the composition of the liabilities. The image below shows [current release](https://www.federalreserve.gov/releases/h41/current/h41.htm) from the Fed on August 28, 2025. ![Screenshot 2025-09-07 at 4.05.53 PM](https://hackmd.io/_uploads/S18dIPsqxg.png) - **The red boxes** show that notes (e.g., paper currency) + deposits (e.g., bank deposits at the Fed itself) account for a whopping 98\% of the \$6.5 trillion of liabilities – pretty wild. - **The green boxes** further break down the deposits into deposits held *by banks* (\$3.2 trillion) versus *by the treasury* (\$595 billion). The treasury account is important because it is how the US government repays the principal of its bonds and is where the proceeds from the sales at the treasury auctions go (more on this later). Given the current composition, I expected that the decline in the aggregate liabilities would be attributed to a decline in either the depository institution deposits or the Treasury general account (I figured that the volume of paper bills wouldn't be able to decrease that quickly – hard to put that genie back in the bottle). Thus, I was a bit surprised to find that it was not the case.[^3] The figure below shows a breakdown of the liabilities over time. [^3]: Granted, those following the Fed's actions more closely over the past years would certainly not be surprised by this. ![liabilities](https://hackmd.io/_uploads/rJsHhDjqex.png =600x) (see the FRED's [BOGMBBM](https://fred.stlouisfed.org/series/BOGMBBM), [CURRCIR](https://fred.stlouisfed.org/series/CURRCIR), & [D2WLTGAL](https://fred.stlouisfed.org/series/D2WLTGAL) series, respectively.) Of note is the fact that none of the balances is at a lower level than in March 2023. In fact, all of them are pretty darn close, in absolute terms, to their previous levels. We can see this even more acutely by pulling all the data into a single plot, as below. ![liabilities](https://hackmd.io/_uploads/HyzP0vo9lg.png) The **bright red** line indicates the sum of the reserves, currency, and Treasury account, while the **dark green** line shows the total liabilities. *While today's liabilities are almost entirely made up of reserves, currency, and the Treasury account, the total liabilities were 2 trillion dollars higher than the sum of the three in 2023.* The missing piece (and thus the namesake of this article) is the [Overnight Reverse Repurchase Agreement](https://fred.stlouisfed.org/series/RRPONTSYD) facility. If reading that set of words makes your eyes glaze over, don't worry; it's actually extremely simple. This object is literally a parking lot for capital. Long story short, the Fed allows people to put dollars here and pays them interest as dictated by the [reverse repo award rate](https://fred.stlouisfed.org/series/RRPONTSYAWARD) (which is moved in lock-step[^4] with the [IORB](https://fred.stlouisfed.org/series/IORB) (the interest on reserves rate, which is what depository institutions earn) and is part of how the Fed implements the Fed funds rate overall). The vast majority of the capital (estimates of 96%) in this facility is money market funds, who aren't depository institutions at the Fed (and thus can't earn the actual interest on reserves rate) but still want to get access to the Fed funds rate, especially when the rates are held at a relatively high rate compared to what treasuries are yielding. This facility has gone from \$2 trillion to about zero in the past two years. The figure below adds this (the balance of the reverse repo) to the composite view of Fed liabilities. [^4]: They move in lock-step, but the IORB is consistently held about 10-15 basis points higher than the reverse repo rate (q.v., currently [4.4\%](https://fred.stlouisfed.org/series/IORB) vs. [4.25\%](https://fred.stlouisfed.org/series/RRPONTSYAWARD), resp.). ![Screenshot 2025-09-07 at 4.59.33 PM](https://hackmd.io/_uploads/rkXnz_ocex.png =400x) The IORB being a bit higher ensures that banks don't prefer using the overnight repo facility to just holding the reserves directly. ![liabilities](https://hackmd.io/_uploads/HJbuXFiqxg.png) (adding in the [RRPONTSYD](https://fred.stlouisfed.org/series/RRPONTSYD) series in cyan.) We now see that the total liabilities are nearly fully accounted for. The story doesn't end here, though, because if you follow the flows *out of* the reverse repo facility, it isn't like that capital evaporates. Given a vast majority of the dollars in the facility were from money market funds, those dollars very likely stay in the Fed orbit because they are just being used to either (i) buy treasuries (e.g., adding to the Treasury general account), or (ii) private repurchase agreements (e.g., with a depository bank and thus increasing that bank's deposit balance with the fed).[^5] Let me restate, because this is the key point: **the massive decline in the balance of the reverse repo facility doesn't, in isolation, account for the massive decline in the Fed liabilities.** The following example can help make this point. [^5]: Tracking the flows of capital that leave the Fed's overnight facility is a harder task because the money market funds can redirect those dollars in many different ways. I justify the claim that "dollars withdrawn from the overnight reverse repo facility remain as Fed liabilities" with two plots. First, we see that the money market funds account for nearly the entire \$2 trillion usage of the facility in the first place. ![Screenshot 2025-09-07 at 6.33.13 PM](https://hackmd.io/_uploads/B1iNOti9xx.png =400x) ([source](https://www.financialresearch.gov/short-term-funding-monitor/market-digests/volume/chart-28/)). Next, we can see that the amount of treasuries in the money market fund holdings increased dramatically over that same time frame. ![Screenshot 2025-09-07 at 6.35.27 PM](https://hackmd.io/_uploads/BJzCdYjceg.png =400x) ([source](https://www.financialresearch.gov/short-term-funding-monitor/market-digests/collateral/chart-2/)). Thus, those dollars are likely refilling the Treasury account or adding to the balances of depository institutions (e.g., staying as Fed liabilities either way). || Reverse repo facility | Treasury account | Fed Liabilities| |-|-|-|-| |Before | \$100 | \$100 |\$200 | | Withdrawal | -\$1 | 0 || | Purchase treasuries | 0 | +\$1 || |After | \$99 | \$101 | \$200 | There are still \$200 of liabilities on the Fed's balance sheet; the money removed from the facility was simply added to the Treasury general account. To square this circle, we need to turn to the asset side of the balance sheet. ## Fed Assets Again, looking at the latest Fed balance sheet [release](https://www.federalreserve.gov/releases/h41/current/h41.htm), we can see the breakdown of assets. ![Screenshot 2025-09-07 at 6.53.43 PM](https://hackmd.io/_uploads/Syj76Yicle.png) As with the liabilities, there are only a few major components that make up the bulk of assets; treasuries (\$4.2 trillion) and mortgage-backed securities[^6] (\$2.1 trillion) constitute 95% of the full \$6.6 trillion in assets. The plot below shows that the asset composition is more stable than the liabilities over time.[^7] ![assets](https://hackmd.io/_uploads/BJI8ycoqex.png =700x) (see [WALCL](https://fred.stlouisfed.org/series/WALCL), [TREAST](https://fred.stlouisfed.org/series/TREAST), [WSHOMCB](https://fred.stlouisfed.org/series/WSHOMCB) for total, treasuries, and mortgage-backed securities time series, respectively.) [^6]: Which weren't on the Fed's balance sheet until 2009, and were the cornerstone purchases of the original quantitative easing in response to the financial crisis. I found it interesting that the Fed didn't just get mortgages on the balance sheet once, but has continually added them in a very similar pattern to treasuries over time. See the [WSHOMCB](https://fred.stlouisfed.org/series/WSHOMCB) series below. ![Screenshot 2025-09-07 at 6.59.20 PM](https://hackmd.io/_uploads/H1wdRYs9gx.png =400x) Someone who knows more about these topics would be better equipped to explain why the Fed chose to continue buying more MBSs. [^7]: As a fun aside, this roughly means that each \$1 bill you have is roughly backed by 64¢ of treasuries and 32¢ of mortgage-backed securities. Funnily enough, the Fed balance sheet lists $11 billion of gold, meaning about 0.167¢ of gold behind each dollar, though sadly, it is not redeemable. So here we are. The assets side of the Fed's balance sheet is shrinking dramatically to implement the latest round of quantitative tightening. Simply, the Fed is both letting the treasuries mature and not reinvesting payments from mortgage-backed securities. When treasuries mature, the principal is repaid *out of the Treasury general account*. This is the connective tissue between how the massive drawdown in the reverse repo balance on the liabilities side is really just being used by the Treasury to repay the principal on treasuries owned *by the Fed itself*. The example below shows this process. | | Treasury account | Reverse repo facility | Fed liabilities | |-|-|-|-| | Before | \$100 | \$100 | \$200 | |At maturity | -\$1 | | | | | \$99 | \$100 | \$199 | |New treasuries | +\$1 | -\$1 || | | \$100 | \$99 | \$199| |At maturity | -\$1 | | | | \$99 | \$99 | \$198 | |New treasuries | +\$1 | -\$1 || | After | \$100 | \$98 | \$198 | In essence, the reverse repo facility is getting drained to refill the Treasury account as new treasuries are being sold. While the number of treasuries that the Fed owns is decreasing, the balance that was held in the reverse repo facility is just getting repurposed into treasuries (see footnote [^5] again).[^8] [^8]: The runoff of mortgage-backed securities is pretty similar. Mortgages are slightly different than bonds because the principal is being repaid gradually with interest rather than at maturity (imagine if instead of monthly payments, a 30-year mortgage meant you had to come up with \$1mm cash in 2065 ... lol). As consumers are making house payments, their checking account balance decreases and pays the owners of the MBS. In this case, the Fed owns the MBS, so the payments are only debited from a depository institution and both sides of the balance sheet contract. ## Summary <u>tldr;</u> **The runoff of treasuries and mortgage-backed securities on the asset side of the Fed balance sheet since March 2023 mainly was absorbed by the gradual flow of capital from the overnight reverse repurchase facility $\rightarrow$ Treasury general account $\rightarrow$ repaying the principal of mature treasuries.** Hopefully, this post helped explain this mechanism a little more clearly. After going through the process of digging through the FRED data, I also found that the discussion of this phenomenon featured prominently in both the 2023 and 2024 annual reports from the NY Fed. > *These dynamics shifted during the second half of 2023, as rates on alternative short-term money market investments began to trade above the ON RRP rate, due to increases in Treasury debt supply and higher demand for repo financing. The slower pace of increases in the FOMC's target range also reduced uncertainty among MMFs regarding the future path of policy, leading MMFs to steadily increase the WAMs of their portfolios from lows of about ten days at the beginning of the year to more than thirty days at year-end. MMFs increased their Treasury bill holdings by $1.20 trillion and their private repo investments by $1.06 trillion in 2023, while corresponding allocations to the ON RRP facility fell by $1.37 trillion over the year to reach the lowest level in more than two years (Chart 4).* > – **Open Market Operations During 2023** [report](https://www.newyorkfed.org/medialibrary/media/markets/omo/omo2023-pdf.pdf); see box 3. Similarly, from 2024. > Federal Reserve liabilities decreased in line with assets during the year, while changing composition. The shift in composition was mainly driven by the sharp decrease in the ON RRP facility from an average level of $807.7 billion in December 2023 to $171.4 billion in December 2024. This decrease reflected broad reallocation by money market fund (MMF) counterparties toward more attractive investments, such as US Treasury bills and private repo. The continuation of strong US Treasury issuance throughout the year supported these shifts in MMF investments. > – **Open Market Operations During 2024** [report](https://www.newyorkfed.org/medialibrary/media/markets/omo/omo2024-pdf.pdf); see "Key Developments." The prominence of the discussion around the reverse repurchase facility validated the importance of the mechanism as it played out over the past two years. Incidentally, in a recent Forward Guidance [episode](https://www.youtube.com/watch?v=P0pIJI4B80g&t=1717s), [Felix](https://x.com/fejau_inc) and [Quinn](https://x.com/qthomp) discussed exactly the primary question that this post raises: *Was the existence of the large reverse repo facility <u>the [crux](https://www.youtube.com/watch?v=ZRTNHDd0gL8&t=236s)</u> of enabling the extended runoff of the Fed's balance sheet?* And if so, what does the fact that the facility is effectively empty mean looking forward? I am not equipped to answer this, but check out the podcast episode for their takes. That's all for now; hopefully, if you read this far, you learned something. For me, it served as a first step into understanding the latest trends in the Fed's balance sheet (and to demystify a few of the terms and FRED time series that get thrown around), but there is so much more to dig into around historical and contemporary central banking. <font color="red">☵ made with ♥ and ✎⭣ by mike ☵</font>