Modelling Canada’s tomorrows

Ever wonder how statisticians project the number of retirees living in Vancouver in 2030 or the number of people working in 2050? Or possible impacts of a tax increase or a policy change?

The answer is simple: We build microsimulation models by using existing data to create communities of virtual people.

Statistics Canada creates models that simulate the lives of real people—not unlike the simulation game SimCity, which teaches players about the inter-relationship among resources. In the game, players manage diverse resources, each of which has an impact on the collective output. In our work, we put our simulated people through real-life paces. Education. Kids. Jobs. Good things happen. Bad things happen.

Approximately 30 Statistics Canada employees are experts in microsimulation techniques. They build models based on simulated lives—individuals and their families—to use for various purposes, from long-term population projections to health analysis to social policy.

Models integrate a huge variety of data to build synthetic, but realistic, individuals. For most of these models, confidentiality is not an issue because the models are based on characteristics of simulated Canadians rather than real Canadians.

Our model citizens in action

“One reason that we do microsimulation is to explore ‛what if’ scenarios. What is the probability of moving? Getting married? Having kids? What is the chance of changing jobs?” says Chantal Hicks, Acting Director of Statistics Canada’s Modelling Division.

Answering such questions is important: they help guide policy-makers deciding where best to put resources. Will their city need more schools, seniors’ residences or immigrants’ services? Will people be able to afford home care and long term care?

“Models, of course, have potential errors. These are projections, not predictions, but the idea is that you can look at the range of outputs that are likely to happen.” Ms. Hicks says.

Microsimulation modelling was first promoted in the 1950s by American economist Guy Orcutt, who advocated using microdata to explain and predict economic behaviour. The techniques became popular in the 1980s, when computers became faster and cheaper to use. In 2009, StatCan hosted the 2nd General Conference of the International Microsimulation Association. Experts from around the world gathered to share techniques made possible by the exponential growth in computer power.

Statistics Canada currently maintains four main policy-relevant models.

The Social Policy Simulation Database and Model, in use since 1988, looks at the possible impact of changing taxes and transfers in Canada by analyzing the financial interactions of governments and individuals. What happens if the GST goes up? What happens if income taxes drop? What is the impact of the federal budget on a single parent or student? Statistics Canada does not analyze policies, but supplies the analytic tool so that others can look at the results when existing tax/transfer programs are modified or to test proposals for new programs.

While the Social Policy Simulation Model explores a point in time, other models follow people through time.

LifePaths offers a look at people from birth to death. It enables users to analyze, develop and cost government programs that have a long-term component, in particular those that need to be evaluated at the individual or family level. LifePaths can also be used to analyze societal issues of a longitudinal nature, such as intergenerational equity or time allocation over entire lifetimes. For instance, researchers have used LifePaths to model pension reform and saving for retirement.

Demosim is used for population projections. Using the microdata file from Canada’s Census of Population as its starting point, Demosim produces dynamic population projections at the level of the provinces, territories, census metropolitan areas and selected smaller geographies.

Population projections can be made for a number of characteristics, such as age, sex, visible minority group, place of birth, generational status, Aboriginal identity, highest level of educational attainment and labour force participation. Demosim does all this by simulating events such as births, deaths, migrations and changes in level of education, according to various population growth scenarios.

Health models are tools, pioneered by Statistics Canada’s Health Analysis Division, to evaluate the impact of health interventions and policies at the population level or to examine the changes in risk factors. What would happen if people exercised more? If screening for colorectal cancer increased? If they chose one treatment option over another? What would be the expected cost of Alzheimer’s disease in 20 years time be if current trends continue?

Drawing from the rich banks of data at Statistics Canada, these microsimulation models realistically represent the Canadian population with attributes, such as risk factor exposures, health histories and typical demographic characteristics. The models simulate histories for individual people in continuous time, and then aggregate individual results to create an outcome for the total population.

Of course, just because things have turned out one way in the past does not guarantee future outcomes. Microsimulation is one tool among many for measuring life in Canada. Nevertheless, our models serve as eyes and ears on the dynamics of Canadian life and give policy-makers a useful bridge between policy and data.

Next month: Assuring quality data