Over the past two decades, federal and state policymakers have dramatically reshaped the nation’s system of cash welfare assistance for low-income families. During this period, there has been considerable variation from state to state in approaches to welfare reform, which are often collectively referred to as “welfare-to-work programs.” This article synthesizes an extraordinary body of evidence: results from 28 benefit-cost studies of welfare-to-work programs based on random assignment evaluation designs. Each of the 28 programs can be viewed as a test of one of six types of welfare reform approaches: mandatory work experience programs, mandatory job-search-first programs, mandatory education-first programs, mandatory mixed-initial-activity programs, earnings supplement programs, and time-limit-mix programs. After describing how benefit-cost studies of welfare-to-work programs are conducted and considering some limitations of these studies, the synthesis addresses such questions as: Which welfare reform program approaches yield a positive return on investments made, from the perspective of program participants and from the perspective of government budgets, and the perspective of society as a whole? Which approaches make program participants better off financially? In which approaches do benefits exceed costs from the government’s point of view? The last two of these questions coincide with the trade-off between reducing dependency on government benefits and ensuring adequate incomes for low-income families. Because the benefit-cost studies examined program effects from the distinct perspectives of government budgets and participants’ incomes separately, they address this trade-off directly. The article thus uses benefit-cost findings to aid in assessing the often complex trade-offs associated with balancing the desire to ensure the poor of adequate incomes and yet encourage self-sufficiency.