Neighborhood Online Networks, Disadvantage and Psychological Distress: Harnessing Big Data on Social Media to Examine Neighborhood Effects

Introduction: Social networks within a neighborhood may protect residents from distress by providing important coping resources. The evidence for the role of social networks in disadvantaged neighborhood has been mixed, with some studies suggesting that networks in these environments cause more distress. The mixed findings may be due to the limited amount of data available to make neighborhood-level inference or the limited attention paid to the role of place-based online networks. The proliferation of smart phones have increased low-income individuals’ access to internet services and online interaction, making it possible to use social media data to infer neighborhood-level distress and networks and improving our understanding of the role of place-based online networks in disadvantaged neighborhoods. This paper, therefore, aims to test the role of place-based online networks in buffering the relationship between neighborhood disadvantage and psychological distress using large-scale social media data and methods of data science.

 

Method: We collected all geotagged tweets (n= 231,302) posted within the physical boundaries of 78 neighborhoods in the City of Pittsburgh during the three-month period (April 1 to June 30, 2013) from the Twitter Streaming API. To measure psychological distress, we employed sentiment analysis using psycho-linguistic lexicon LIWC (Linguistic Inquiry and Word Count), i.e., anger, anxiety, and sadness. Social networks represent the strength of social connections maintained over social media between neighborhoods. We quantified the social network strength between two neighborhoods based on the number of replies sent between users’ home neighborhoods. Neighborhood social networks were computed as the proportion of ties of a neighborhood received from within it. Higher within-networks suggest more social networks received within a neighborhood. Neighborhood poverty was measured using neighborhood-level rates of poverty and unemployment; and disorder was measured using crime rates, tax delinquency and vacant lots. These measures were then dichotomized based on mean scores.  

 

Results: Among 78 Pittsburgh neighborhoods, 35 neighborhoods were characterized as high-poverty neighborhoods and 43 as low-poverty neighborhoods. Compared to low-poverty neighborhoods, high-poverty neighborhoods had significantly higher levels of neighborhood disorder [t = 2.2 (73.18), p < .05], and significantly lower levels of psychological distress [t = -2.59 (58.36), p < .001], sadness [t = -2.59 (58.36), p < .001], and anxiety [t = -2.59 (58.36), p < .001]. More online social networks within a neighborhood was significantly associated with the increased of psychological distress [β = 0.004 (0.001), p < .01], sadness [β = .004(.001), p < .001], and anxiety, [β = .003(.001), p < .01] in high-poverty neighborhoods. However, this pattern was not found in low-poverty neighborhoods.

Implication: Similar to social networks that exist offline, place-based online social networks have a negative rather than supportive effect on distress in disadvantaged neighborhoods. Online social networks may be targeted to decrease distress in disadvantaged neighborhood, but more needs to be known about the contents of online social networks and how they can be modified to be more supportive. Leveraging large-scale data analytics on social media can allow researchers to improve upon the existing methods to understand the relationships between neighborhoods and psychological distress.