Household and Job Relocation

The household and job relocation models in POLARIS capture long-term location decisions — specifically, the probability that a household moves to a new home or that a person changes their work location over a given time horizon. These models are relevant for multi-year forecasting scenarios where demographic and employment changes affect travel demand patterns.

Modeling Approach

Both models use a hazard-based duration model (also known as a survival model) to estimate the expected tenure at the current home or job. The hazard function determines the probability that a transition (relocation) occurs within a given timeframe, conditional on the current tenure and a set of explanatory variables.

The figure below illustrates the parallel decision flows for home and job relocation, showing how household- and person-level attributes feed into the hazard models, which determine whether a relocation occurs, triggering the location choice model to assign a new zone.

The general form of the linear predictor is:

\[\beta \mathbf{x} = \text{ASC} + \sum_k \beta_k \cdot x_k\]

where $\mathbf{x}$ is the vector of explanatory variables and $\beta$ are the estimated coefficients. The baseline hazard follows a log-normal specification with a scale parameter $\sigma$.

Home Relocation Model

The home relocation model operates at the household level and estimates the probability that a household changes its residential location. Explanatory variables include:

Variable	Description
Home ownership	Whether the household owns or rents
Housing type	Apartment vs. single-family home
Number of rooms	Small dwelling indicator (1-2 rooms)
Household income	Income at the time of moving in
Number of children	Household members under 18
Number of females	Female household members
Structural changes	Whether the household moved without changes in composition
Rent payment	Whether the household pays rent
Number of students	Primary and tertiary student counts
Primary driving mode	Whether driving is the primary commute mode

When the model determines that a relocation occurs, the household is reassigned a new home location using the same location choice model as the initial population synthesis.

Job Relocation Model

The job relocation model operates at the person level and estimates the probability that a worker changes their job location. Explanatory variables include:

Variable	Description
Income	Final income in thousands of dollars
Employment type	Full-time work indicator
Work flexibility	Flexible work arrangement indicator
Work from home	Telecommuting indicator
Number of jobs	Multiple job holder
Occupation	Administrative, manager, or professional indicators
Demographics	Gender, age

When a job relocation is triggered, the person is reassigned a new work location through the routine activity location choice process.

Application in Multi-Year Scenarios

In POLARIS, these models are invoked when running multi-year simulations (e.g., evaluating the impacts of demographic shifts or land-use changes over a 5-20 year horizon). For each simulated year, every household and worker is evaluated:

1. The expected tenure is estimated using the hazard model.

2. A stochastic draw determines whether a relocation event occurs within the forecast timeframe.

3. If relocation occurs, the location choice models are re-invoked for the affected household or person.

This process ensures that the synthetic population evolves realistically over time, capturing the churn in residential and employment locations that drives changes in commute patterns and travel demand.