A growing number of studies have examined increases in “deep” or “extreme” poverty among families with children and trends in the number of disconnected single mothers. Most of these studies use Census survey data with varying amounts of underreporting of safety net benefits. Our study contributes to this literature by correcting for underreported TANF, SSI, and SNAP income in the Current Population Survey using baseline data from the ASPE/Urban Institute TRIM3 microsimulation model. These corrections provide a more accurate picture of the economic well-being of families with children and the impact of the safety net on their poverty levels. Our data include both cash and near-cash income and cover the 1995-2010 period, allowing us to analyze the impact of both welfare reform and the recent recession.  Our corrections substantially lower the measured levels of deep poverty in the CPS in any given year.  Perhaps surprisingly, however, we find that the corrections reveal greater increases in deep poverty among children over the decade 1995-2005, while showing less of an increase in poverty and deep poverty during the 2005-2010 period, relative to the uncorrected data.  Consistent with past SIPP analyses – but unlike uncorrected CPS data – the corrected CPS data show that the share of Americans below half the poverty line increased significantly in the decade before the Great Recession, and that this was due to the diminishing poverty effects of income-support programs, especially TANF. Another finding is that safety net and recovery programs responded effectively during the Great Recession, largely averting what would have been a far larger surge in poverty and deep poverty. The results shed light on the strengths and weaknesses of a work-based safety net in a time of high joblessness, as well as the continuing importance of temporary Recovery Act provisions. Our paper also provides detailed characteristics of who is in deep poverty -- and for which groups deep poverty has grown the most – including data on age, race, ethnicity, disability, work participation, benefits receipt, marital status, immigration status and family structure.