We develop a model where CLOs are the optimal financial structure for securitizing assets that trade in illiquid markets. As is true empirically, CLOs hold portfolios of risky loans that they actively rebalance and finance themselves with safe, long-maturity debt. CLOs face restrictions on their equity payouts, but these restrictions ignore the market-to-market value of CLO assets. Banks that hold CLOs’ safe debt are insulated from fire-sales, and a financial system where CLOs take risk instead of banks is more efficient than traditional banking. CLOs are most useful in moderately illiquid markets with significant gains from trade and fire-sale costs.

We use bond-level data to study how US banks manage their securities portfolios, focusing on bank responses to the rapid shift in interest rates in 2022-23, and examine the role of financial and regulatory frictions in shaping bank behavior. Interest rate risk in bank portfolios increased sharply as rates rose, but with significant cross-bank heterogeneity depending on the ex ante share of bonds with embedded options. Exposed banks did not, however, offset the rise in risk either by selling long-duration bonds or hedging using "qualified'' accounting hedges. We identify two frictions that may help account for this inertia. First, we show banks are highly averse to selling underwater bonds at a discount to book value---e.g., banks were 8-9 times more likely to trade bonds with unrealized gains than unrealized losses in 2022-23. This "strategic'' trading is more pronounced for banks that do not recognize unrealized losses in regulatory capital and banks with low stock market valuations. Second, frictions in establishing qualified accounting hedges limited hedging activity depending on bond type and accounting classification. Banks did, however, reduce the sensitivity of regulatory capital to interest rates by classifying the riskiest bonds as held-to-maturity.

Greater reliance on nonbank financing makes firms fragile as it leads banks to limit their access to credit lines. Besides demonstrating this result in panel tests subject to range of controls and robustness checks, we employ the 2014–16 oil-price collapse as an exogenous rollover risk in nonbank financing of non-oil-sector firms by collateralized loan obligations (CLOs) exposed to the oil sector. Nonbank-reliant firms with looming maturities or higher credit risk face reductions and wider spreads in bank credit lines after the shock, resulting in weaker financial and real performance in spite of their drawdowns of existing credit lines.