What the Data Say About Repo Sensitivity

This chapter interprets the Federal Reserve note’s evidence on daily repo-rate movements: what the regression measures, how Treasury bills, coupons, dealer holdings, and MMF assets enter the specification, and why the estimated response changes with liquidity relative to GDP. The practical goal is to read the numbers as conditional evidence about repo capacity—not as a universal formula that turns any Treasury issuance or reserve movement into a predetermined rate change.

The distinction matters because the balance-sheet mechanism supplies a story, while the data tell us how strongly that story appears in a particular sample and specification. A good reading must keep three layers separate. The first is the reported observation: an estimated association in the data. The second is the economic interpretation: which borrower or lender balance sheet may be responding. The third is the boundary: what the regression cannot establish about causality, other market segments, or future episodes.

The regression is a map, not a meter

The note estimates daily changes in the spread between the tri-party general collateral rate, TGCR, and the interest rate on reserve balances, IORB. Both rates are measured in basis points, so the dependent variable asks whether the repo rate is moving relative to the rate associated with bank reserve holdings. The sample runs from September 2014 through March 2026 and contains 2,902 observations.

The explanatory variables represent different channels. Net Treasury bill issuance and net Treasury coupon issuance are daily measures in billions of dollars. The specification also includes lagged changes in primary-dealer Treasury holdings and lagged changes in government MMF assets under management, or AUM. Those two series are weekly and are interpolated to daily values. Federal Reserve liquidity is measured as reserves plus ON RRP balances, scaled by GDP. The specification includes lagged changes in the repo spread as well, allowing the estimate to account for some persistence and reversal in daily movements. Standard errors are clustered by week to account for serial correlation.

Scaling liquidity by nominal GDP gives the comparison a denominator. A dollar amount that might be manageable in a large economy can represent a much tighter condition in a smaller one, so the note compares Fed-provided liquidity with the size of the economy rather than treating the raw balance as context-free. That normalization does not mean GDP itself is a direct cause of the repo rate. It is a way of comparing the financial system’s liquidity environment across time and across different absolute balance-sheet sizes.

That list is more than statistical housekeeping. Each variable is a different answer to the question, “What changed at the margin?” Bill issuance changes the supply of an asset that MMFs may prefer to hold. Coupon issuance can increase the quantity of longer-dated Treasury securities that dealers and leveraged investors need to finance. Dealer holdings approximate the amount of Treasury inventory on a major borrowing side of the market. MMF AUM approximates the potential scale of a major lending side. Reserves plus ON RRP describe the aggregate liquidity supplied through the Federal Reserve, but not how much of it is offered in repo.

The reported magnitudes make the units concrete. In the note’s specification, $100 billion of net Treasury bill issuance is associated with a 1.3-basis-point increase in repo rates. The corresponding association for $100 billion of net Treasury coupon issuance is 3.9 basis points.