THE CENSUS BUREAU reported last week that the poverty rate rose in 2003 for the third consecutive year and that the percentage of people without health insurance climbed for a third straight year as well. But what's really revealing about the trends in poverty and health coverage is not the similarity, which reflects the downturn in the business cycle, but rather the contrast. Put simply, the long-term trend in health coverage is negative: The percentage of Americans without insurance is up from about 13 percent in the late 1980s to nearly 16 percent today. But the long-term trend in poverty is positive: The rate ranged between 13 percent and 15 percent in the 1980s, but it has hovered at about 12 percent in the past half-decade. Moreover, the real decline in poverty is bigger than suggested by these numbers.

What policy prescriptions follow? The health insurance figures are a call to action: Economic growth over the past 15 years has failed to improve coverage, so some kind of policy response is essential. Indeed, this conclusion is only reinforced if you dig into the data. The percentage of uninsured has gone up despite expanded government activism: Since 1987, the first year for which the Census Bureau has data, government has gone from covering 23.3 percent of the population to covering 26.6 percent, a rise that primarily reflects the extension of Medicaid to a wider circle of low-income people. But the trend in company-provided health insurance has been negative, except during the extraordinary bubble of the late 1990s. The soaring cost of health insurance is driving companies to withdraw coverage, and there's little reason to expect this trend to reverse itself. More government activism -- possibly in the form of incentives to companies to keep offering insurance, such as those proposed by Sen. John F. Kerry -- seems unavoidable.

Interpreting the poverty data is harder. Economic growth has been a powerful antidote to poverty, and the long-term reduction in the poverty rate is all the more impressive because it defies powerful head winds. High immigration in the past two decades might have been expected to increase poverty; the same goes for the rise of single-parent families, which are four to five times as likely to be poor than two-parent ones. Yet economic growth, while powerful, is not a magic bullet. For one thing, the surprisingly small increase in the poverty rate during the recent economic downturn owes something to the policies of the 1990s, which provided financial incentives, training, child care and other services to people trying to escape poverty via work. For another, the positive news on the number of people in poverty masks an alarming trend. The number of people in extreme poverty -- that is, subsisting on less than half the income defined as the poverty line -- stands at 15.3 million, higher than at any time since the Census Bureau began collecting data 28 years ago.

This fact deserves more attention than it has received. The policies of the 1990s may have successfully pushed former welfare recipients into work, but those who have not found work, or who have found it and then lost it, appear to be worse off than before. A bit oddly, the congressional debate on reauthorizing the welfare law has dwelt minutely on the details of the design of work incentives rather than on the program's function as a safety net; it is not often noted, for example, that fewer than half the families eligible for money and job training actually receive any, a shocking contrast with the participation rate of 80 percent that prevailed in the mid-1990s. A combination of deft government incentives and a strong economy may have reduced overall poverty. But the people at the very bottom are being forgotten.