The data collected across 92 buildings were used to develop a model of Parking Utilization (observed parked cars per occupied housing unit in the building) at the parcel level. The independent variables were chosen to optimize both the model’s goodness of fit and predictability. The tested variables were grouped into two major categories, variables that describe the building and those that describe the surrounding neighborhood. Variables that describe the building include:

- Parking Supply – Number of stalls provided divided by the total number of units in the building
- Transit Information – Variable equals 1 if transit information is available for building tenants
- Fraction of Affordable Units – Fraction of units set aside for affordable housing
- Average Unit Size – Average square feet for all units in the building
- Parking Price – The average price charged for parking one car in the buildings parking facility
- Average Bedrooms per Unit – Average number of bedrooms per unit for all units in the building
- Average Rent – Average rent for all units in the building

Variables that describe the neighborhood include:

- Block size (walkability measure) – Average size of all blocks that intersect a ¼ mile buffer around each parcel
- Retail & Service Job Density (retail proximity measure) - The number of employees working in these establishments was totaled for establishments within ¼ mile of the parcel. This total is then divided by the land area within this the ¼ mile area.
- Walkable Transit Trips per Day (access to transit measure) – Number of trips available within a ¼ mile for buses and ½ mile for rail using network distances, divided by the area (in acres) within a ¼ mile of the parcel.
- Jobs by 45 Minute Transit (job accessibility by transit measure) – The transit commute time is determined from every block in DC to every Transportation Analysis Zone (TAZ). The numbers of jobs (from the Metropolitan Washington Council of Governments) in the TAZs that are within a 45 transit trip are totaled to create this measure.

Development of the regression model considered interactions between the independent variables. For example the Transit Trips per Hour variable was correlated with Parking Utilization, but once walkability (measured by Block Size) and all the other variables were introduced into the regression it was found that the statistical significance was reduced to a level that would not include it in the final model. However, if Transit Trips per Hour and Block Size were interacted then the interaction variable was found to meet the significance criteria. Using this flexible model form has the advantage of finding significant combinations of independent variables; however, it does make the model somewhat more complicated to understand. The final model is presented below:

Variable 1 | Variable 2 | Incremental R^{2} |
---|---|---|

Parking Supply per Unit | -- | 65.7% |

Block Size | -- | 70.1% |

Transit Information | Walkable Transit Trips/Day | 72.3% |

Fraction Affordable Units | Jobs by 45 Minute Transit | 74.3% |

Sq. Ft. per Unit | Jobs by 45 Minute Transit | 76.3% |

Parking Price | Retail/Service Job Density | 77.8% |

Block Size | Walkable Transit Trips/Day | 78.1% |

Average Bedroom/Unit | Block Size | 78.4% |

Average Bedroom/Unit | Sq. Ft. per Units | 79.9% |

Average Rent | Retail/Service Job Density | 80.2% |

Retail/Service Job Density | Parking Supply per Unit | 82.5% |

Generally, as parking supply, average unit size, average number of bedrooms, average rent, and block size increases (or walkability decreases), parking utilization increases. Where there is provision of transit information, greater walk access to transit, greater transit access to jobs, and higher retail and service job density, parking utilization decreases.

DDOT, DCOP, and the consulting team are developing a peer reviewed paper detailing the entire technical process to be posted to this site at a later date.