Every leveraged position in the world sits on top of a market almost nobody outside of fixed income thinks about, and I include my past self in that. For years I treated overnight funding as a solved problem, a utility that just runs in the background like electricity. Then September 2019 happened, the overnight repo rate briefly spiked to something like four times where it should have been, the Fed had to jump in with emergency operations, and I finally sat down to understand the thing I had been ignoring. What I found is that the repo market is both simpler and more fragile than its reputation suggests, and once you understand the fragility you get a genuinely useful early-warning signal for risk assets.
What a repo actually is
Strip away the jargon and a repurchase agreement is a collateralized loan that lasts about a day. I have Treasuries, I need cash, so I sell you my Treasuries tonight and agree to buy them back tomorrow at a very slightly higher price. That tiny price difference is the interest. You get a safe overnight place to park cash and you are holding my collateral the whole time, so if I vanish you just keep the bonds. That is the entire trick. It is how banks, hedge funds, money market funds, and dealers move cash and securities around each night to square their books.
Two details do most of the work in a stress episode. The first is the collateral itself. Treasuries are the gold standard because everyone agrees on what they are worth, so they fund cheaply. Push into less pristine collateral and lenders get pickier. The second is the haircut, which is the buffer the lender demands. If I want cash against a bond, you might only lend me ninety-eight cents on the dollar of its market value. That two percent gap protects you if the bond falls in price before I buy it back. Haircuts feel boring right up until they move, and when they move they move fast, because a rising haircut forces the borrower to find cash from somewhere else at exactly the moment cash is scarce.
Why it seizes up
The thing that made September 2019 click for me is that the repo market has a hard capacity limit that has nothing to do with how much cash exists in the system. That limit is dealer balance sheet. When a hedge fund needs to fund a Treasury position overnight, it usually goes through a bank dealer who intermediates the trade. The dealer takes the collateral on one side and passes cash on the other, and both of those legs consume balance sheet capacity that regulations cap. So even when there is plenty of money in the banking system in aggregate, the pipes that carry it can clog because the intermediaries have no room left.
In September 2019 a couple of ordinary things landed on the same day. Corporate tax payments drained cash out of the system, and a batch of Treasury settlements pulled more cash toward the government. Reserves that had already been drifting lower were suddenly stretched thin, and dealers who might normally have stepped in to lend against Treasuries were near their balance sheet limits and simply did not. Rates that should sit near the Fed's target gapped higher for a day because the demand for overnight cash met a supply of intermediation that could not flex. Nothing was insolvent. The plumbing just did not have the width to handle the surge, and the price of overnight cash exploded until the Fed showed up and provided it directly.
That is the mental model I keep now. A funding squeeze is rarely a story about credit or solvency. It is usually a story about too much collateral chasing too little balance sheet on a specific date, and the calendar tells you when those dates are likely to cluster.
Which rates to actually watch
You do not need a terminal to track funding stress. A handful of published overnight rates carry most of the signal, and the useful information is in the spreads between them, not the levels.
- SOFR, the Secured Overnight Financing Rate, is the broad measure of what it costs to borrow cash against Treasuries overnight. It is the headline gauge. When SOFR jumps well above where the Fed wants short rates to sit, funding is tightening.
- SOFR versus the interest rate the Fed pays on reserves. When secured overnight borrowing costs more than the risk-free rate banks earn just for parking money at the Fed, that is a sign cash is genuinely scarce in the plumbing, not just repriced.
- The spread of SOFR to the effective fed funds rate. Fed funds is unsecured, SOFR is secured against the safest collateral there is. In calm times secured should not cost meaningfully more than unsecured. When it does, something in the collateral or balance sheet channel is jammed.
- The percentiles inside SOFR. The rate is a distribution, and the top percentile prints move first. If the ninety-ninth percentile is running hot while the average looks fine, stress is building at the edges before it hits the headline.
My rule of thumb is that persistent pressure matters and one-day spikes usually do not. A single quarter-end or tax-date pop that mean-reverts the next morning is the calendar doing its thing. A spread that stays elevated for several sessions and keeps grinding wider is the market telling you intermediation capacity is running out, and that is worth acting on.
What it historically meant for risk assets
Here is the part that matters if you never plan to touch a Treasury. Funding stress tends to lead risk-off in other markets rather than coincide with it, because it hits leverage directly. When overnight cash gets expensive or hard to source, the first thing that happens is that levered players get squeezed on their carry. Positions that were comfortable at cheap funding become uncomfortable, and the natural response is to trim. That deleveraging can show up as weakness in equities and credit and crypto in the days and weeks after the funding signal, not the same afternoon.
I want to be honest about the limits of this. A repo spike is not a reliable timing tool, and treating it like one will get you chopped up. September 2019 is instructive precisely because the Fed intervened and risk assets kept climbing afterward, so the naive reading of that single episode is wrong. The signal is not that funding stress guarantees a selloff. It is that funding stress removes a support that leverage quietly depended on, and it tells you the plumbing is close to its limit so the next shock lands on a system with less give in it.
The practical version I use is a short checklist. When a funding spread pushes wide and stays there for more than a session or two, I do three things. I check whether it lines up with a known calendar date, because if it does the pressure probably fades on its own. I look at whether reserves in the system have been trending down over months, because a thin system reacts harder to the same shock. And I lighten up on the most leverage-sensitive corners of my book rather than trying to call a top, because the honest read of a funding signal is that the odds of a sharp move went up, not that a crash starts tomorrow. Most of the time nothing dramatic follows, and that is fine. The point is not to predict the break. It is to notice when the thing everyone leans on is running out of room, and to be carrying a little less when it does.