New government health insurance programs are likely to emphasize voluntary purchases in a market setting, with subsidies targeted at low-income populations and stress on managed care. Such programs are best structured with a guaranteed enrollment period that is as long as six months to a year. However, given that incomes change over time, errors will be made in awarding income-related subsidies for that long. These errors are assessed in simulations undertaken with longitudinal data from the Survey of Income and Program Participation. Two allocations of the subsidies, based on current income at the beginning of the enrollment period and on actual income assessed at the end, are compared for a variety of program designs. Prospective determination of subsidies is somewhat biased toward overpayment. Net overpayments amount to 5-10 percent of subsidy costs. However, prospective payment encourages participation in the subsidy program. The simulated participation rate for true eligibles is as high as 73 percent with prospective subsidies, compared to 69 percent with retrospective reconciliation. Net overpayments are slightly reduced by testing income less frequently and over longer periods.