Evaluation of E-Scooter Media Coverage

Findings This study examines the attitudes and common themes of 840 e-scooter news articles from across the US, published between 2017 to 2020. Sentiment analysis suggests that media coverage does not appear to be predominantly in favor or against shared e-scooters. A word count analysis shows that the most discussed themes were conflicts with other road users, safety concerns, and regulation efforts. The findings reveal differences between regions but also common patterns, such as the decline in negative coverage with time.

. Newspaper articles used in this study by region and publication.

Methods
This study uses quantitative content analysis methods to examine online local and national news articles discussing shared e-scooters. We collected 840 escooter-related online articles from ten US metropolitan areas and separately at the national level (Table 1). We employ sentiment analysis to flag each article's attitude, and we use word count to identify the most common issues discussed by the media.
2. What are the issues discussed in the media coverage of e-scooters?
3. How did e-scooter media coverage change over time?

Do trends in e-scooter media coverage differ between regions?
Evaluation of E-Scooter Media Coverage

Findings
In the statistical software R, we used the Tidy Text package (Silge and Robinson 2016), which breaks down the articles into words, removes connecting words (conjunctions, prepositions, and others), and stemming the words (e.g., "injury" and "injuries" become "injuri").
Using the "Google News Search" engine, we searched for articles from September 2017 to August 2020 -a period of three years since the first introduction of shared e-scooters, with the keywords "scooters" and "scooters city name." We only selected news articles from known local and national newspapers and media networks and well-known blogs.
We used a dictionary-based approach to sentiment analysis. The affiliation of sentiments is based on pre-prepared unigram dictionaries (Hardeniya and Borikar 2016) -Bing (Jindal and Liu 2008), Afinn (Nielsen 2011), and NRC (Mohammad and Turney 2013). In the classification process, the algorithm counts the appearance of affiliated words in each document, and the output includes the number of appearances of each sentiment (Priyavrat and Sharma 2018;Silge and Robinson 2016). Next, we summed the number of negative and positive sentiment words in each article and summed the negative/positive word balance for each dictionary. We assigned '-1' for negative balance, '0' for equal balance, and '1' for positive balance. The sign of the average for each article provided the final sentiment.
Finally, we counted the words in each examined region and calculated the percent of each word in the corpus. We chose the twenty most common words that can provide insights into the issues discussed in the news articles and grouped them into themes. These themes represent the main issues addressed in the articles. The introduction of e-scooters in a region was mostly accompanied by positive coverage. In some regions, negative coverage followed, possibly as a reaction to e-scooter's externalities. Notwithstanding, San Francisco's and Tampa's reporters showed an immediate negative response. Interestingly, the number of negative e-scooter articles across the US and in most cities sharply diminished in 2020. That may be linked to the eruption of the COVID-19 pandemic. This trend may suggest the acceptance and the normalization of shared e-scooters by the media. For the word counts, we categorized the twenty most common words into three themes (Table 2): Conflicts -conflicts between e-scooter users and operators to others, primarily pedestrians. Regulation -efforts to organize and regulate e-scooter usage and operation in cities. Safety -safety concerns and health hazards associated with e-scooter users and their surroundings.

Findings
The share of these themes in the corpus changed between regions over time ( Figure 2). However, safety was the most discussed issue across the entire sample, followed by regulation and conflicts between pedestrians and escooters. In most of the regions, conflicts were highly prevalent in media accounts following the introduction of e-scooter services. The national perspective shows a slight move from discussing conflicts to discussing safety, which may present an evolution from concerns to actual incidents.

Evaluation of E-Scooter Media Coverage
Findings Table 2. Share of the 20 most frequent words resembling e-scooter issues across the sampled articles. The letters (C), (R), and (S), represent the themes.  Figure 2. Quarterly share of the three main issues discussed in e-scooter media by region.
* The Y-axis range varies in the different charts.
Generally, the three subjects were highly discussed throughout the three years.
In some cities, one of the issues is less concerned than other issues, such as conflicts in Baltimore and regulation in Portland. That tells about the local experience of each of these regions with the adoption of e-scooters. Sharp peaks in the coverage of one issue in different regions mostly correspond with a sharp increase in negative coverage in that region, sometimes because of a new regulation. For example, eight articles covering the ban of e-scooters from sidewalks in August 2019, five articles covering Uber's e-scooters suspension in Los Angeles in November 2019, and five articles covering San Diego's e-scooter boardwalk ban in May 2018.
In addition to the continuous expansion of e-scooter services, our findings suggest that shared e-scooters could become a part of cities' landscape, finding their niche in urban mobility and being normalized by the public and the media.

Evaluation of E-Scooter Media Coverage
Findings