Abstracts

Mechanical, calendar-driven trades (target-date rebalancing, option expiry hedging, dividend reinvestment) are individually small but collide. The birthday paradox implies collision counts grow quadratically in the number of mechanical traders, so nonfundamental price pressure scales with n², not n, even as each trader's footprint shrinks. We model markets as a combinatorial collision process and show fat-tailed returns and volatility clustering emerge without behavioral assumptions or leverage. The model predicts volatility spikes on ex-ante identifiable "collision-dense" dates and that indexing growth raises tail risk quadratically. Evidence from fund-flow calendars and expiry schedules supports both predictions.

Most prices in the American bond market were never paid by anyone. On a typical day more than four fifths of corporate and municipal securities do not trade, yet each closes at a mark supplied by one of two valuation firms, and these manufactured prices, model outputs rather than transactions, govern fund net asset values, collateral calls, and insurers' regulatory capital. Exploiting quasi-random reassignment of identical securities across pricing vendors by fund administrators, we show manufactured prices behave like administered prices: fundamental news enters marks with a five-week half-life, and the largest repricing events follow vendor methodology revisions, one of which moved $1.4 trillion of reported wealth overnight. Smoothness is the equilibrium: administrators pay for pricing errors but never for staleness, vendors compete accordingly, and the speed at which the financial system recognizes losses becomes an industrial choice, risk arriving on balance sheets slowly, then all at once.

Intermediary asset pricing ties risk premia to intermediary equity, but equity omits the sector's largest asset: the deposit franchise. We estimate franchise capital, the present value of rents from sticky, below-market deposit funding, using a structural demand system for deposits, and find it averages 1.3 times the book equity of U.S. banks. A pricing kernel built on franchise-inclusive capital doubles the cross-sectional explanatory power of equity-based kernels and resolves the 2022 puzzle in which marked-to-market bank equity fell by a fifth while risk premia barely moved: rising rates raised franchise capital, leaving total capital nearly unchanged. Franchise capital is run-contingent, however; it exists only while depositors stay. Accounting for this, the sector's risk-bearing capacity in run states is 40% smaller than equity implies. Intermediaries appear best capitalized exactly when their capital is most fragile.

Asset pricing explains countercyclical risk premia with representative agents whose risk aversion swings over the business cycle. We show no agent's preferences need move: the identity of the marginal investor does. Using investor-level demand elasticities identified from quasi-random fund flow shocks, we estimate each investor class's state-contingent pricing weight, a measure we call SDF incidence. In calm markets, elastic long-horizon institutions price equities; in downturns, pricing migrates to constrained broker-dealers and leveraged households, whose steeper demand curves require 40% higher expected returns for the same cash flows. This rotation accounts for two-thirds of the predictable variation in the equity premium and reproduces the value and volatility premia without habit, long-run risk, or disasters. Discount rates vary because the economy's pricing kernel changes hands, not because anyone's risk appetite does.

Since 1993, essentially the entire U.S. equity premium has accrued between the close and the open; intraday returns average zero. We show this pattern is the price of overnight inventory risk charged by constrained intermediaries, not news timing or retail demand. Dealers cannot rebalance overnight and risk limits bind on close-of-day positions, so absorbing closing-auction order imbalances requires a return concession that reverses at the open. Three results support the mechanism. First, exploiting the staggered rollout of overnight trading sessions on alternative trading systems between 2021 and 2025, stocks gaining access to overnight liquidity lose 58 percent of their overnight premium relative to matched controls. Second, the premium doubles in the week before quarter-end regulatory reporting dates, but only in stocks intermediated by dealers whose regulators compute leverage ratios from quarter-end snapshots rather than daily averages. Third, the premium loads on intermediary capital shocks, concentrates in stocks with the largest closing imbalances, and is absent where liquid 24-hour futures permit hedging. A calibrated inventory model matches both the level and the pre-snapshot spikes. Much of the equity premium is rent paid to balance sheets that warehouse risk while markets sleep.