We show that the answer lies in commitment: a threat works only if the dealmaker will actually carry it out. Seniority can credibly punish holdout creditors, while similar dilution is not credible in takeovers, making cash the natural solution. More commitment is not always better—partial increases can actually make holdouts worse.

Bankruptcy is costly, while out-of-court restructuring is often blocked by creditor holdouts. We show that reducing bankruptcy costs need not crowd out restructuring; instead, it can facilitate restructuring. Making bankruptcy more debtor-friendly can have a similar effect. We also show how bankruptcy law shapes creditors’ willingness to restructure, when government support can backfire, and why controversial liability-management exercises such as uptiers and dropdowns may be less consequential than commonly thought—or even irrelevant.

Not if debt can be diluted. We show that long-term interbank debt lets distressed banks borrow against claims on healthy counterparties, turning interconnectedness into private liquidity insurance—a “backdoor bail-in” without regulators or contingent securities. Policies that compress exposures or restrict senior emergency borrowing can therefore make the system look safer while weakening its ability to absorb shocks.

We show that full insurance can encourage firms to keep bad projects alive, so cash requirements work like a copay: they put the firm’s own money at risk when it draws. This also explains a striking pattern—firms with low pledgeability need liquidity insurance most, yet receive less of it and face tighter covenants and greater revocation risk.

Firms do not choose foreign markets in isolation: each investment reshapes the map of attractive future destinations. Using Chinese multinationals, we show that firms expand toward countries that are geographically close to—and strongly connected with—their existing foreign operations. Global expansion is therefore path dependent: where a firm has already invested helps determine where it goes next.