Conventional-politics approaches, emphasizing party ideology, electoral dynamics, committee membership, campaign donations, and industry clout, exercise a powerful hold over assessments of public policies and their distributional effects. Emerging from pluralist and business power perspectives, such accounts see “who gets what and why” as the result of how politics and power shape policies, their implementation, and distributional outcomes. This pervades our understanding of the Paycheck Protection Program (PPP), the US government’s effort to avert mass unemployment during the COVID-19 pandemic by lending $786 billion to small businesses to keep employees on the payroll. Yet contrary to prior studies of the PPP, we find that conventional-politics factors were strikingly uncorrelated with distributional outcomes, revealing limits to such approaches to this case. Instead, we find that an institutional politics or politics-in-time (IP-PIT) analysis better explains the program and its trajectories. IP-PIT revises the causal sequence by emphasizing how institutions and policies generate politics, distributional outcomes, and feedback loops. We engage both approaches via a mixed-methods analysis of the PPP and two new datasets. We conduct a qualitative process tracing of temporal variation in policy architectures, politics, policy revisions, and access to loans across the program’s three periods, and present quantitative analyses of loan flows across congressional districts and periods using data on the entire corpus of PPP loans. In so doing, we advance research and debate over the PPP, the dynamics and outcomes of US policy making during crises, and the American political economy in general. Ours is the first study of the PPP to conduct a mixed-methods analysis of loans across congressional districts or to use conventional and institutional approaches to address its politics, policy, and outcomes. More broadly, we document varieties of critical junctures, contribute arguments about what might shape policy or institutional innovation in those moments, and use the PPP to identify conditions under which systems are “their own grave diggers,” fueling negative-transformative rather than positive-reproductive feedback.