http://senseable.mit.edu/papers/pdf/20210130_Salazar-etal_DesirableStreets_CEUS.pdf
Conclusions
This paper leverages the revealed preference of thousands of pedestrians deviating from their shortest path to construct a desirability index for streets. This measure captures the scenic and experience value provided by different parts of the city. Beyond its conceptual value, the desirability index offers practical value for urban planners by allowing them to identify areas with high potential to attract pedestrians. Conversely, it can also be used to pinpoint areas with low desirability that would be suitable candidates for street improvements and other interventions.
The first part of the paper describes how we construct the desirability index and then shows its application to Boston. We show that people systematically deviate from the shortest path to frequent highly desirable streets. This pattern is consistent across different types of trips, including trips taking place at different times of the day (morning, noon, afternoon, and night trips) and trips of different lengths (defined by the total length of the shortest path). This finding suggests that although in principle, people may value the features of the built environment differently depending on the time of the day and the length of the trip, the desirability index captures a common set of features that are desirable to most pedestrians in our sample. Second, we show that desirable street segments are widely dispersed throughout the study area. From this observation, we learn that there are streets across all neighborhoods whose characteristics appeal to pedestrians. Moreover, these local differences in the built environment are commonplace, and pedestrians across all neighborhoods systematically deviate from their shortest path to experience them. The geographic dispersion of desirable streets also highlights a methodological advantage of using deviations from the shortest path rather than the count of pedestrians to describe the desirability of streets. Since the construction of the desirability index exploits very local differences in the environment instead of differences across large geographic areas, it allows us to characterize desirable streets relative to their nearby context instead of comparing broader differences between neighborhoods—for example, by comparing streets located at the center versus the periphery. Another advantage of this approach is that it allows us to trace how specific streets change in time. This approach can be useful, for example, to revitalize distressed streets or detect areas affected by blight or other negative phases of urban change.
In the second part of the paper, we analyze a diverse set of built environment characteristics that have been shown to affect the pedestrian experience. In particular, we construct a set of measures that aim to capture the provision of amenities in each street segment, as well as some of its urban design features. To do so, we combine detailed data on businesses, amenities and the street layout pulled from OpenStreetMaps with urban design characteristics of streets collected from Google Street View imagery. By comparing street segments with varying levels of desirability and different built environment characteristics, we document that the desirability index is systematically associated with the provision of public amenities. More desirable streets are characterized by having better access to parks, sidewalks, and a higher presence of urban furniture. These results provide complementary evidence underscoring the importance of pedestrian infrastructure for supporting pedestrian activity (Saelens & Handy, 2008a, 2008b), and also strengthens evidence on the importance of parks for physical activity and leisure (Coombes, Jones, & Hillsdon, 2010; Lee & Maheswaran, 2011).
The paper also provides evidence on the urban design characteristics that are associated with more desirable streets. In particular, desirable streets tend to be more visually enclosed, sinuous, and tend to have more homogeneous facades. Our results on visual enclosure align with normative urban design proposals suggesting that great walking environments should feel like a grand corridor with buildings that define and enclose space (Jacobs & Appleyard, 1987; Speck, 2012). The fact that desirable streets are more sinuous aligns with the view that irregular street forms can easily accommodate focal locations for landmark buildings and reduce traffic due to their curvature. Finally, our findings on desirable streets being more homogeneous in terms of their facade present empirical support for the notion that architectural homogeneity is positively valued by pedestrians, which aligns with existing evidence on residential property markets (Lindenthal, 2020), evidence pointing to commonalities in architectural preferences (Nasar, 1992), and more broadly with studies underscoring the importance of urban design characteristics and aesthetics in determining behavior (Kopec, 2012; Lynch, 1960). Finally, our results point to desirable streets having more business establishments. Although the number of business establishments is an important determinant of desirability, the variety of businesses is more important in predicting it. In addition, we find that among all the included businesses, retail is consistently associated with higher desirability of streets. These results are in line with theories that suggest that the density and variety of land uses can encourage frequent opportunities for interaction (Jacobs, 1961; Talen, 1999), can raise the vitality of places (Yue et al., 2017), and also aligns with evidence suggesting that retail shops are key to increase pedestrian travel (Lund, 2003).
There are several limitations in this study that could be addressed in future research. First, due to a non-disclosure agreement, we have very little information about the individuals whose devices we observe in our data. Not having access to demographic data, for example, limits the generalizability of our findings. Second, the information we do have comes from an unknown selection process based on when devices’ phone applications request and log location data. Although the data generating process is uncertain, the granularity of our data (pings at 1 second intervals) is frequent enough to confidently provide a detailed picture of where people are at different types of the day. Third, because the GPS data is collected as people go about their daily life, it tends to overweight commuting patterns. This limits how much we can say about the type of people being attracted to specific streets and the types of activities they engage with once they are there. Finally, the desirability measure we propose is a local measure. This means that although it can be calculated for entire cities in practice, the desirability of a street should always be interpreted with respect to its immediate context. The local nature of the measure limits the extent to which one can draw useful comparisons about the average desirability of large geographical areas or cities.