This section is intended to outline the process by which the analysts addressed their research questions [see Inquiry]. It provides details regarding sourcing, analytics tools, data validation, and calculations performed. [To view a complete list of sources used, refer to Sources.]
Building permit application data was obtained in June 2020 from the City of Portland, Oregon, using the website PortlandMaps. This application dataset ranged from Jan. 2000 to June 2020, covering approximately two full real estate cycles.
Land use review application data was also obtained in June 2020 from the City of Portland, Oregon, using the website PortlandMaps. This application dataset ranged from Jan. 2010 to June 2020, covering approximately one full real estate cycle. For this application dataset, the following columns had to be transcribed from PortlandMaps: address, description, new units or lots, issue date, final date, and completed date.
Transportation system data was obtained in July 2020 from the City of Portland, Oregon, using the website PortlandMaps Open Data. This dataset contains the ID as well as the latitude and longitude coordinates of 70 active public transit stations, including 69 MAX light rail stations and 1 streetcar/tram station.
Demographic datasets for the City of Portland were sourced from the U.S. Census Bureau. Demographic maps in the form of GeoJSON files were sourced from the City of Portland, Oregon, using the website PortlandMaps Open Data.
Topics of inquiry were sourced from qualitative interviews with industry professionals regarding trends of interest in Portland’s real estate landscape.
Tools Used for Analysis.
Datasets were stored and organized using Microsoft Excel. Addresses for each building permit application and land use review application were geocoded using Google’s Geocoding API, chosen for its comprehensiveness and relative accuracy. Datasets were analyzed and visualized using kepler.gl, an open-source geospatial analytics tool created by Uber; Chart Studio, a cloud-based data analytics platform created by Plotly; R, a programming language; and Stata, a data science software package.
Data Validation, Quality Assurance, & Completeness.
The following records were filtered out of the building permit application dataset: redundant records, including DFS (deferred submittal) and REV (renovation) permit applications; applications that added no new units; and duplicate records.
The following were filtered out of the land use review application dataset: applications that added no new units or lots, and duplicate records. Note that because the land use review application dataset had to be partially transcribed by hand, there is some potential for increased error.
The building permit application dataset was cross-verified with the U.S. Census Bureau reports for yearly housing production in the City of Portland. The land use review application dataset was partially cross-verified with an independently transcribed dataset provided by real estate developer Noel Johnson.
The transportation system dataset was independently validated by real estate developer Noel Johnson. Non-active stations were filtered out.
Time, in days, until city approval and completion was calculated for every application that contained the requisite information. [See section IV for information regarding approval and completion time formulas.]
Addresses for all applications, if available, were geocoded to latitude and longitude coordinates using Google’s Geocoding API. Distance to the city core was calculated, with the city core defined as the Pioneer Courthouse, located at 700 SW 6th Ave, Portland, Oregon. Using the transportation system dataset, the nearest public transit station for each application was identified, and the distances were calculated. [See section IV for information regarding distance formulas.]
Calculations & Formulas.
City approval time, in days, for both building permit applications and land use review applications were calculated as follows:
t = Date - Date
Time until project completion for both types of applications were calculated as follows:
t = Date - Date
Note that Datefinal refers to either the final date listed or the date of latest activity.
Distances, in miles, between sets of latitude and longitude coordinates were calculated using great-circle distance formulae adapted for Excel notation.