A Correlational Study of Microblogging Patterns and Population Adaptive Capacity amid a U.S. Natural Disaster

During a disaster, the resilience of a community can rapidly change as the community copes with the disaster, this is a measure of its adaptive capacity; therefore, disaster managers need a better way of determining the adaptive capacity of the affected community as the event evolves. Microblogging adds a dynamic element to adaptive capacity, which is based on the types of messages the affected population posts as the disaster evolves. Understanding how social media is used to cope with a disaster may help disaster managers identify topics of concern to the affected population and to develop more effective relief and recovery plans. A quantitative correlational study was conducted to examine whether relationships exist between the Twitter message posting patterns and the social vulnerability of a population affected by a major U.S. natural disaster. Five coastal counties impacted by Hurricane Sandy were selected. Six variables were used: Five predictor variables were informational-related messages (INF), action-related messages (ACT), opinion-related messages (OPI), emotion-related messages (EMO), and general disaster-related information messages (GEN) and one criterion variable, the Social Vulnerability Index (SoVI). The final sample size for the study was 75 message samples, created from the sampling frame of 11,474 messages. Hypothesis 1 results showed 14 significant correlated pairs with weak negative correlations between INF and SoVI, ACT and SoVI, OPI and SoVI, and GEN and SoVI, and 10 positive intercorrelations among the predictor variables (p< .05). EMO was a significant positive predictor of SoVI and GEN was a significant negative predictor of SoVI (p<.05). Hypothesis 2 results showed the five predictor variables (INF, ACT, EMO, OPI, and GEN) collectively contributed to explain 37.7% of the variance of SoVI. The key recommendations for future research included (a) a quantitative correlation study to examine predictive relationships among message classification variables and SoVI time of recovery, (b) an inferential regression study to examine the predictive relationships among message classification variables and census tract SoVI data, and (c) inferential regression study to improve the accuracy of the sample messages.