Crisis information distribution on Twitter: a content analysis of tweets during Hurricane Sandy

Abstract

Social media has been widely used for crisis communication during disasters, and its use during extreme events has drawn attention from both researchers and practitioners. Although crisis information coverage and distribution speed are important issues, both have not been studied extensively in the literature. This paper fills this gap by studying information distribution and coverage of social media during disasters. To this end, we searched and analyzed 986,579 tweets posted during Hurricane Sandy (October 22 to November 6, 2012). To learn about responses from official agents, we sampled 163 governmental organizations (GO), 31 non-governmental organizations (NGO) and 276 news agent accounts and their tweets for analysis. Specifically, five social media key performance indicators (KPIs) are studied in this paper, including impression, like, mention, re-tweet, and response time, and other variables such as hashtag, tweet frequency, and information type. We also test whether the five KPIs and other variables are different among different user types. Results show that total impression, re-tweet rate, hashtag, and tweet frequency are significantly $(P<0.05)$ different among different user types. Specifically, although news agent users generate a larger number of total impressions and tweet more frequently than GO and NGO users, their re-tweet rates and number of hashtags are lower than the GO and NGO users. Re-tweet rate based on mentioned users (5%) is significantly higher $(P=0.00)$ than that based on regular followers (0.01%). Nearly 89% of total impressions are generated from regular followers, with impressions from re-tweeting being a minority. This paper provides some new insights into how social media was used for crisis communication during disasters.