A Census Post-Enumeration Survey (PES) is conducted after each Australian Population Census. The PES provides a measure of the net under-count of the Census, and is a key input into the production of the official Estimated Resident Population counts.

This paper describes estimation of net under-count in the context of the 2006 Census and PES. It introduces a new estimator, the prediction regression (PREG) estimator, as an extension of the dual system estimator (DSE) standardly used in estimating Census under-count. In contrast to the DSE, the PREG estimator can use a variety of benchmark variables without the need to form non-overlapping post-strata. It can also adjust appropriately for persons that report different categories in the PES than were recorded in the Census.