The world around us is not naturally organized into categories for statistical analysis. For the purposes of data journalism, discrete, unique incidents, events, and people must be rendered as similar, so that abstract categories may be created and compared, a new study states.

Wilson Lowrey and Jue Hou, of the University of Alabama, studied data journalism, focusing on the  construction of abstract categories. They did a content analysis of 194 data journalism projects from 2011 to 2016. These were made by legacy and non-legacy outlets in the United States and United Kingdom.

The results show increasing abstraction and decreasing attention to personal, lived anecdotes.

Abstract constructs were evident in the form of calculated metrics and aggregated indices that were both adopted and invented by the data project producer. These were found in 36% of the cases. The use of anecdotal reporting decreased from 44% to 26% over the years.

There was little evidence of easy-to-find data scrutiny reporting, for example mentioning limitations in data sampling. About one-third of the projects mentioned such issues.

Government sources were dominant, making up almost half of all the sources used. During 2011-2016, the amount of government sources decreased and self-gathered data were used more. The use of national and pan-national sources outweighted the use of local sources. Large organizations are most likely to produce complex data sets. Howevere, locally sourced data was not decreasing over the years.

Statistical categories and metrics are alluring, the researhers remind. “They are, however, social, political, and economic constructs and not objective reflections of the world”, they conclude.

The article “All forest, no trees?” was published in Journalism and is available on the publisher’s website (abstract free).

Picture: Abstract fluid art by Lurm, license CC0 1.0