This paper uses a life cycle model to study interactions between household self-insurance and the U.S. Disability Insurance (DI) system. The model is motivated and guided by evidence from panel data on disability onset in U.S. households, showing that married workers benefit from both higher self-insurance capacity and higher utilization of DI compared to unmarried workers---who are left, by contrast, more exposed to the costs of disability. These responses are consistent with adverse selection, whereby the long application process and strict work limitations of the DI system screen out worse self-insured workers. Accounting for household self-insurance and the implicit costs of utilizing the DI system, the model delivers novel insights into the welfare implications of DI reform. Welfare gains from DI reforms are large, especially ones that lower the costs of acquiring DI benefits and consequently provide income support to households that value it highly. Accounting for the substantial insurance value that expansionary reforms provide is important for drawing these welfare conclusions. On the other hand, accounting for the self-insurance provided by spousal labor supply and pooled family savings is also important, as it reduces welfare gains from DI reforms by as much as 25 percent.