Employment Relocation Model

Employment relocation and location choices are made by firms. However, in the current version of UrbanSim, we use individual jobs as the units of analysis. This is equivalent to assuming that businesses are making individual choices about the location of each job, and are not constrained to moving an entire establishment.

The Employment Relocation Model predicts the probability that jobs of each type will move from their current location or stay during a particular year. This is a transitional change that could reflect job turnover by employees, layoffs, business relocations or closures. Similar to the economic transition model when handling job losses in declining sectors, the model assumes that the hazard of moving is proportional to the spatial distribution of jobs in the sector. All placement of jobs is managed through the employment location model.

As in the case of job losses predicted in the economic transition component, the application of this model requires subtracting jobs by sector from the buildings they currently occupy, and the updating of the accounting to make this space available as vacant space. These counts will be added to the unallocated new jobs by sector calculated in the economic transition model. The combination of new and moving jobs serve as a pool to be located in the employment location choice model. Vacancy of nonresidential space will be updated, making space available for allocation in the employment location choice model.

Since it is possible that the relative attractiveness of commercial space in other locations when compared with an establishment's current location may influence its decision to move, an alternative structure for the relocation model could use the marginal choice in a nested logit model with a conditional choice of location. In this way, the model would use information about the relative utility of alternative locations compared to the utility of the current location in predicting whether jobs will move. While this might be more theoretically appealing than the specification given, it is generally not supported by the data available for calibration. Instead, the relocation decision is treated as an independent choice, and the probabilities estimated by annual relocation rates directly observed over a recent period for each sector. The resulting form of the employment relocation model is as follows.

$R_{st}$ is a set of jobs that are chosen to be moved based on $D(j, t)$ , a binary decision function based on Monte Carlo sampling process using the annual relocation rate for sector $s$ .

$$R_{st} = \{\, j \in J_{st} \mid D(j,t) \,\},$$ (8)

where:

$R_{st}$	is the set of jobs in sector $s$ at time $t$ that are uprooted by the relocation model,
$D(j, t)$	is a function based on Monte Carlo sampling process determining if job $j$ will be moved at time $t$ .

The jobs to be moved are now unplaced and added to the unplaced jobs set, and the space they occupied will be released and become available to unplaced jobs for location choice.

$$J^U_{st} = J^U_{st} \cup R_{st}$$ (9)

$$S_{lt} = S_{l(t-1)} + s^l_j \mid j \in R_{st}, (j, l) \in M^l_{jt... ...\{\, (j, l) \vert j \in R_{st}, (j, l) \in M^l_{jt} \}\, \Delta J_{st} < 0, \\$$ (10)