SPENSER
SPENSER (Synthetic Population Estimation and Scenario Projection Model) is an open-source population microsimulation framework developed by the University of Leeds.
| Developed by | University of Leeds |
| Website | lida.leeds.ac.uk/spenser |
| Licence | Open source |
| Purpose | Generate and project high-resolution synthetic populations |
It is designed to generate high-resolution synthetic representations of households and individuals, and to project these populations forward under alternative demographic and socioeconomic scenarios.
Individuals, not totals
Rather than forecasting population only as aggregate totals, SPENSER represents individual people and households with demographic and socioeconomic characteristics, and simulates how their composition and spatial distribution change over time.
The framework combines two techniques:
| Technique | What it contributes |
|---|---|
| Spatial microsimulation | Constructs a population whose small-area characteristics match observed constraints |
| Dynamic microsimulation | Advances that population through time as individuals age and their circumstances change |
Combining the two allows user-defined assumptions concerning economic conditions, policy, health, housing or other future changes to be translated into demographic processes such as fertility, mortality, household change and migration.
SPENSER therefore provides a detailed population basis for infrastructure, land-use, health and transport-demand modelling, particularly where individual- or household-level population information is required.
Required inputs
SPENSER requires base-year demographic and household information from which a synthetic population can be constructed, together with demographic transition rates and future scenario assumptions.
Census data and microdata
Two different roles, often confused:
| Source | Role |
|---|---|
| Census data | Provides constraints on the number and characteristics of individuals and households within small geographical areas |
| Census microdata or survey data | Provides relationships among characteristics that cannot be obtained from aggregate tables alone |
Why both are needed
Aggregate census tables tell you how many people in an area are over 65, and separately how many live alone. They do not tell you how many over-65s live alone, because that cross-tabulation is suppressed or unpublished.
Microdata supplies those joint relationships, which are then applied within the constraints the aggregate tables set.
Additional demographic inputs
- ONS population estimates and projections
- Household projections
- Fertility and mortality rates
- Internal and international migration rates
- Information describing household formation and changes in household composition
Longitudinal data for transition probabilities
Longitudinal datasets such as Understanding Society can be used to estimate transition probabilities for events including:
- Ageing
- Partnership formation and dissolution
- Childbirth
- Changes in employment status
- Household mobility
- Migration
Spatial and socioeconomic variables
When constructing user-defined scenarios or modelling residential migration, additional local characteristics can be incorporated:
- Housing availability
- Employment
- Crime
- Green space
- Other local characteristics
Generated outputs
SPENSER generates spatially detailed synthetic populations in which individual persons are assigned to households and characterised by demographic and socioeconomic attributes.
It can project these individuals and households through time using dynamic microsimulation, producing future populations under a baseline projection and alternative user-defined scenarios.
Outputs may include:
| Output | Level |
|---|---|
| Population by age, sex and other characteristics | Individual |
| Household size and composition | Household |
| Socioeconomic and employment characteristics | Individual |
| Residential location | Individual and household |
| Corresponding small-area population and household totals | Aggregate |
Why the individual-level output matters
Because the outputs retain individual- and household-level information rather than only aggregate population counts, they can be used directly as inputs to downstream agent-based and travel-demand models, with no disaggregation step in between.
For transport-demand modelling, SPENSER can therefore provide the future synthetic population from which traveller agents are constructed and subsequently assigned activity patterns, workplaces, destinations, modes and travel schedules.
The University of Leeds specifically notes that SPENSER outputs have been used as inputs to transport and land-use models, illustrating its role as a population-generation and scenario-projection component within wider integrated modelling frameworks.