Unemployment Insurance (UI) benefits were a central part of the social safety net during the COVID-19 recession. UI benefits, however, are severely understated in surveys. Using administrative tax data, we find that over half of UI benefits were missed in major survey data, with a greater understatement among low-income workers. As a result, 2020 official poverty rates were overstated by about 2 percentage points, and corrected poverty reached a six-decade low. We provide data to correct underreporting in surveys and show that, compared to UI benefits, the UI exclusion tax expenditure was less targeted at low incomes.