Evaluating binary alignment methods in microsimulation models
Jinjing Li & Cathal O'Donoghue
#2012-003
Alignment is a widely adopted technique in the field of microsimulation
for social and economic policy research. However, limited research has
been devoted to the understanding of their simulation properties. This
paper discusses and evaluates six common alignment algorithms used in
the dynamic microsimulation through a set of theoretical and statistical
criteria proposed in the earlier literature (e.g. Morrison 2006;
O'Donoghue 2010). This paper presents and compares the alignment
processes, probability transformations, and the statistical properties
of alignment outputs in transparent and controlled setups with both
synthetic and real life dataset (LII). The result suggests that there is
no single best method for all simulation scenarios. Instead, the choice
of alignment method might need to be adapted to the assumptions and
requirements in a specific project.
Key words: alignment, microsimulation, algorithm evaluation
JEL Classification: C1 C5