An empirical total survey error decomposition using data combination

Survey error is known to be pervasive and bias even simple, but important, estimates of means, rates, totals, such as the poverty unemployment rate. In order summarize analyze extent, sources, consequences survey error, we define empirical counterparts key components Total Error Framework that can estimated using data combination. Specifically, estimate total decompose it into three high level sources error: generalized coverage item non-response measurement error. We further these lower failure report a positive amount errors in amounts conditional on reporting value. For dollars paid by two large government transfer programs, use administrative records universe program payments New York State linked major household surveys previously defined. find varies its size composition, always far largest source Our application shows combination makes possible routinely measure components. results allow producers assess reduction strategies users mitigate or gauge reliability their conclusions.