Abstract:
|
Existing estimates of changes in poverty are misleading as they rely on household surveys and income concepts that lead to systematic biases. Household surveys suffer from pronounced income misreporting that has substantially increased over time. Income concepts typically exclude some components of income such as in-kind benefits, tax credits, or the rental equivalent of car and home ownership. We examine changes in poverty that address these issues using administrative program and tax data linked to Census Bureau surveys starting in 1995. We rely on the Comprehensive Income Dataset (CID), which links household survey data with an unprecedented set of administrative tax records and program data that cover most major transfer programs. In doing so, we provide the first-ever evaluation using linked survey and administrative data of changes in poverty over time in the United States. We also estimate changes in deep poverty (below 50% over the poverty line) and near poverty (below 150% over the poverty line). Finally, we compare our results using the linked survey and administrative data to results based on other data sources including survey data alone and consumption data.
|