Objectives
Dilapidated or neglected aspects of the built environment — such as abandoned houses or litter — are recognized across disciplines as harmful to community wellbeing and safety. Researchers utilize a variety of tools to observe these items, with Google Street View (GSV) being among the most recent advances. While GSV allows researchers to virtually visit locations without the time, cost, and resources required to go in-person, the validity and reliability of these observations may raise concerns.
Methods
In this study, we examine inter-method and inter-coder reliability for 22 items consistent with established indicators of physical disorder and guardianship. We compare physical disorder observations across 5 researchers who coded 30 streets in Baltimore City, both in-person and using GSV.
Results
Using the intra-class correlation coefficient, we find higher reliability across coders and methods for items that are less transitory (e.g., boarded up windows). In addition, consistency issues between GSV and in-person methods were prevalent and largely pointed to validity issues in GSV. Specifically, many items that were observed in person were “missed” in GSV.
Conclusions
While our findings indicate a preference for the validity and reliability of in-person observation, they do not serve as a final verdict on the utility of GSV.