Abstract:

Many researchers consider the presentation ofdiverse content as a prerequisite for the newsmedia to fully exercise their democratic mandate.While prior news diversity studies have contributedimportant theoretical insights, we argue here thatscholarly knowledge of this concept can be significantlyadvanced by employing computational methodsfor text analysis. Using automated methods,researchers can increase both the scope of databeing analyzed and the resolution of the analysis.This article presents a novel framework for analyzingnews diversity consisting of two distinct stages. Inthe first stage, a computational text classificationmethod is used to analyze, at a high resolution, theattention given in news texts to a broad range ofpolitical and social issues. In the second stage, thetext classifications are aggregated, and the distributionsof media attention to those issues (i.e., newsdiversity) are assessed on a large scale. After presentingthe novel approach, we illustrate its usefulnessfor testing theoretical hypotheses about news diversity.We compare the diversity of economic coveragein three elite and three popular US newspapers(N = 252,807 articles) and find that a fine-grainedanalysis relaxes concerns raised in previous studiesabout low content diversity in the popular press.