No abstract

Background Social distancing is an important component of the response to the COVID-19 pandemic. Minimizing social interactions and travel reduces the rate at which the infection spreads and “flattens the curve” so that the medical system is better equipped to treat infected individuals. However, it remains unclear how the public will respond to these policies as the pandemic continues. Objective The aim of this study is to present the Twitter Social Mobility Index, a measure of social distancing and travel derived from Twitter data. We used public geolocated Twitter data to measure how much users travel in a given week. Methods We collected 469,669,925 tweets geotagged in the United States from January 1, 2019, to April 27, 2020. We analyzed the aggregated mobility variance of a total of 3,768,959 Twitter users at the city and state level from the start of the COVID-19 pandemic. Results We found a large reduction (61.83%) in travel in the United States after the implementation of social distancing policies. However, the variance by state was high, ranging from 38.54% to 76.80%. The eight states that had not issued statewide social distancing orders as of the start of April ranked poorly in terms of travel reduction: Arkansas (45), Iowa (37), Nebraska (35), North Dakota (22), South Carolina (38), South Dakota (46), Oklahoma (50), Utah (14), and Wyoming (53). We are presenting our findings on the internet and will continue to update our analysis during the pandemic. Conclusions We observed larger travel reductions in states that were early adopters of social distancing policies and smaller changes in states without such policies. The results were also consistent with those based on other mobility data to a certain extent. Therefore, geolocated tweets are an effective way to track social distancing practices using a public resource, and this tracking may be useful as part of ongoing pandemic response planning.

Computational social science studies often contextualize content analysis within standard demographics.Since demographics are unavailable on many social media platforms (e.g. Twitter), numerous studies have inferred demographics automatically. Despite many studies presenting proof-of-concept inference of race and ethnicity, training of practical systems remains elusive since there are few annotated datasets. Existing datasets are small, inaccurate, or fail to cover the four most common racial and ethnic groups in the United States. We present a method to identify self-reports of race and ethnicity from Twitter profile descriptions. Despite the noise of automated supervision, our self-report datasets enable improvements in classification performance on gold standard self-report survey data. The result is a reproducible method for creating large-scale training resources for race and ethnicity. 1 https://bitbucket.org/mdredze/demographer 2 http://www.cs.jhu.edu/ ~mdredze/demographics-training-data/

No abstract