ABSTRACT
Measuring visual preferences for urban landscapes is of interest to several fields, and this chapter demonstrates how this type of research can be transformed by using new data sources and methods. Based on a large sample of street voting data from a crowdsourced online visual preference survey , the authors apply the Elo rating algorithm to calculate beauty scores and regresses on these scores with urban design indicators constructed using GIS data. The best fitting OLS model finds that eight variables can explain roughly 43% of the variation in the beauty score and the most significant predictors are the number of trees and the number of intersections. Other significant predictors include street length, properties with windows, proportion of build land, park and open space land, historic buildings, active uses, number of buildings and bike racks. While consistent with previous research on urban landscape preferences, the demonstrated method has applications in fields of urban design, landscape evaluation, real estate and urban economics. The chapter shows that while online methods can successfully reach large numbers for landscape preference surveys, these datasets will require new analysis techniques to obtain findings from such big data.
