Page 297 - Invited Paper Session (IPS) - Volume 1

IPS153 John D.

Estimation of gross population flows when SPDs
are incomplete
John Dunne
Central Statistics Office, Ireland

Abstract
Gross population flows are made up of population inflows and population
out¬flows in the population stocks between two time points or over a period
of time. Population inflows comprise of immigration and births while
population outflows comprise of emigration and deaths. A Statistical
Population Dataset (SPD) is a dataset that typically contains one record for
each member in the population. The compilation of an SPD is typically the first
step in the compilation of population estimates from microdata. If a traditional
Census based population estimate, the final census dataset can be considered
an SPD. With a register-based population estimate, the SPD is based on the
population register. SPDs can also be compiled directly from administrative
data as in the Irish PECADO (Population Estimates Compiled from
Administrative Data Only) project. When SPDs are complete and assuming
perfect linkage, gross population flows can simply be derived by compar¬ing
the SPDs between two consecutive timepoints or periods. Population
estimates are simply obtained by taking counts from the SPD. If the SPD
contains only undercount and not overcount then the SPD (list A) can be
adjusted using a second list (list B) in a Dual System Estimation (DSE) setup to
obtain population estimates. The PECADO project proposes a system of
population estimates where both lists are compiled from administrative data
sources. Reliable population estimates will provide for reliable estimates of net
population flows but not gross population flows. We review the system of
population estimates in the PECADO project before proposing an extension
of DSE methods for the estimation of gross population flows that relies on the
same lists used to compile population estimates. While not perfect, the
proposed method has the property that as the coverage of the SPDs becomes
complete the bias and variance of the gross population flow estimates goes
to zero. We also consider alternative options and argue against them in favour
of the proposed method. Finally, we present estimates of gross population
flows using the proposed method and compare them to the official estimates.
The practical application in the PECADO project focuses on two consecutive
periods (calendar years) rather than timepoints.

Keywords
administrative data; DSE; migration; undercoverage; SPD; PECADO

286 | I S I W S C 2 0 1 9