A Match Made In Silicon: Marriage Matching Algorithms for Dynamic Microsimulation
part of the DYNACAN project, a full-fledged Canadian actuarial stochastic model designed for the fiscal and policy analysis of social security schemes.
DYNACAN was a Canadian, stochastic (Monte Carlo), open, longitudinal, dynamic microsimulation model developed for generating longitudinal and cross-sectional, as well as fiscal and policy-oriented analyses of Canadian Social Security Schemes such as the Canada Pension Plan (CPP), the Old Age Security (OAS) Program and the Guaranteed Income Supplement (GIS).
Marriage and Remarriage: This module had two sub-components.
First, each simulation year, it identifies appropriate numbers of persons, separately by gender, as being "in the marriage market." The probability of entering the pool depends on a person's characteristics, e.g. age, education, earnings level, marital status (never-married, separated, divorced, widow(er)ed) and the presence of dependent children. Second, using a compatibility index based on the individuals' characteristics, the module serves as "matchmaker" to create
specific couples from the gender-specific pools of individuals in the market that simulation year. We align marriage and remarriage rates using Statistics Canada vital statistics information on the age and gender-specific fractions of people marrying in a given year, and some U.S. data from CORSIM where Canadian data are not available.
The compatibility measure employed by DYNACAN is a logit equation that includes the difference in the individuals' ages and its square, the difference in their years of education, the number of children the woman has, race, labour force participation, and the difference in their earnings. The equations also calculate various interaction effects between variables, such as the male's education if he is the older of the two, and the product of the female's education and the male's earnings.
It is defunct now, and I think it was because the compatibility equation DOES NOT include personality traits, nor any kind of "similarity" calculation in personality traits.