This paper studies market depth in the world's most traded equity derivative, S&P 500 E-mini futures. Using a previously unused feature of a public data feed, we count at millisecond resolution the number of agents posting limit orders, a new measure of liquidity supply that we use alongside standard depth. Market depth has fallen by 80% since 2012, driven by the withdrawal of liquidity providers (LPs) rather than a reduction in per-LP capacity. To shed light on this finding, we develop a model of endogenous liquidity provision in which LPs face participation costs to enter and risk-bearing constraints once active. The model delivers testable predictions linking active LP participation to volatility, illiquidity, and LP risk-bearing ability in line with the data: LP participation rises and then falls with volatility and illiquidity, a pattern that distinguishes endogenous entry from a fixed market-maker population, and tighter constraints reduce LP numbers. Conventional price-based liquidity measures have not deteriorated alongside depth, but with eighty percent fewer LPs absorbing risk the market's buffer for the next stress event has thinned.

Interdealer bond markets are highly concentrated, yet prices often appear competitive. We reconcile this paradox by showing that repeated interactions across multiple bonds—multimarket contact (MMC)—discipline dealers’ pricing behaviour. MMC leads to lower interdealer spreads with direct pass-through to end investors. This contrasts with evidence from industries such as airlines and banking, where greater MMC fosters tacit collusion and higher markups. In a dynamic setting where dealers are tempted to widen spreads, we demonstrate that a higher contact between two dealers induces tighter spreads of all dealers intermediating the same bond. Using regulatory TRACE data, we confirm the model's predictions and find that dealers with greater MMC quote tighter spreads in both interdealer and customer trades. Finally, exploiting the post-COVID issuance boom and pre-COVID dealer specialization, we identify the causal effect of MMC on spreads via a shift-share instrument. Our findings show that cross-market relationships among intermediaries—often viewed as a source of collusion—can instead make prices in concentrated markets more competitive.

We propose a dynamic model of bidding in treasury auctions, in which primary dealers must satisfy minimum winning requirements to retain their dealer status. Data from Argentina between 1996 and 2001, a period in which primary dealer requirements were particularly important, shows dealers bid more aggressively the greater their shortfalls in meeting the requirements, thus sacrificing short-term profits to retain their status. We then leverage this trade-off and develop a method for estimating the value of being a primary dealer. We estimate that the gain from being a dealer is of the same order of magnitude as short-term profits. Dealers who bid optimally retain dealer status with high probability, but may have to sacrifice a significant amount of short-term profits to do so. Finally, we use our model to perform a counterfactual exercise which illustrates how the central bank can use minimum winning requirements in order to reduce dealers’ rents.