Introduction

This study intends to examine the relationship between rapid transit and property values in the City of Vancouver, British Columbia.  There have been numerous other studies looking at the topic of proximity to transit and its effect on property values, with mixed results on its effects. Most literature, however, have used American cities as case studies, and almost none have looked at Vancouver.

In Greater Vancouver, the construction of rapid transit takes the form of what is known locally as SkyTrain, an elevated network of automated trains which cross the region.  The Expo line was constructed, as the name suggests, in time for the 1986 World Expo, followed by the Millennium Line, which finished construction in 2002.  Most recently, the Canada Line opened in 2009 for the Winter Olympics, connecting the airport to Downtown.

With over 400,000 daily boardings on the Skytrain per day (Translink, 2011), the public transit system has played an important role in moving residents and visitors around the Lower Mainland.  20% of the region’s population take public transit to commute to work (Statistics Canada, 2011).  Rapid transit has also influenced the way people work and live in the city.  For instance, the location of stations has triggered (re)development of adjacent land in the form of new employment offices and residential buildings.  Some comprehensive planning efforts have also contributed to forms of transit-oriented development (TOD) particularly around Joyce-Collingwood station in the City of Vancouver.  Further (controversial) developments are underway around other stations such as Oakridge and Marine Drive.

Using data derived from Statistics Canada, I select a set of socio-economic and neighbourhood-level variables to estimate property value, and then introduce a distance measure to the closest Skytrain station to see if this affects price.  The assumption is that properties close to SkyTrain stops will show a premium in prices compared to areas further away.  A multivariate regression model will be used to predict this, followed by a geographically weighted regression to look at more local variation in the results.