BBC online local news analysis
One key aspect of the UK Government Mid-Term Review of the BBC Charter is examining how the BBC and Ofcom assess the market impact and public value of the BBC in an evolving marketplace and how that relates to the wider UK media ecology, including with regards to commercial radio, local news, and other content producers and distributors.
In its six-year blueprint, The BBC Across the UK published in 2021, the BBC lays out ambitious objectives to transform the future portfolio of BBC local services. The goal is to “touch more lives, in more communities, more often”, covering the stories that matter most to local audiences and representing different voices and perspectives.
Building on the six-year blueprint, in an announcement made in October 2022, the BBC also committed to (i) creating 11 investigative reporting teams across the country, (ii) implementing multimedia news operations across radio, TV and online local news and (iii) boosting online local news provision by reinvesting in online and multimedia production. The broader aim of this is to “reshape [the BBC’s] local services across England to deliver greater online impact and provide more original journalism”.
However, the BBC’s strategy has come under criticism from news media industry stakeholders, who have expressed concern that increasing the BBC’s focus on local news would draw away readers from established commercial local news providers and potentially put local publishers out of business.
Previous research (KPMG 2015) found limited evidence that the BBC had either a significant negative market impact on, or crowded out, commercial news providers during the previous Charter period. However, since the last Charter review in 2016, the acute financial challenges faced by news providers, particularly at the local level, have worsened. As highlighted by the 2019 Cairncross Review, local news providers have been particularly affected by the shift from print to online news consumption, a transition which has accelerated as a result of the Covid-19 pandemic.
In light of these shifts in the UK news media sector, to shed further light on the relationship between the BBC and commercial providers in online local news provision, our research team developed an innovative, large-scale quantitative methodology centred on collecting a broad range of local news articles across the UK. We focused on three specific research questions:
What are the differences in the content and type of local news coverage provided by the BBC and the commercial sector?
What volume of local news events are covered by both commercial news providers and the BBC?
Where the BBC and the commercial sector cover the same events, are there differences in how these events are reported?
Our first step was to build a comprehensive dataset of UK local news articles published online by the BBC and commercial providers. We built this dataset using a four-step process:
We developed a list of UK commercial local news titles and their website URLs, using data from the Audit Bureau of Circulations (ABC) and the Joint Industry Currency for Regional Media Research (JICREG).
We identified all articles published by each title across a variety of databases (News API, Bing Web and Bing News, Twitter accounts of local news providers) using an automated approach.
We developed a definition of local news and kept only relevant articles in our dataset.
We identified commercial articles and BBC articles reporting on the same local news events using search engine results (sorted by relevance).
This dataset allowed us to calculate quantitative metrics at the article level, including total number of words, sources, quotes, hyperlinks, references, images and similarities in language and themes/topics (drawing on natural language processing models for textual analysis). These metrics meant we could assess the scope of content duplication between the BBC and commercial providers and whether it was consistent with multiple providers covering the same news events, requesting quotes from the same sources and describing the facts or key features of the news events in a consistent manner.
For the final step, we needed to match BBC and commercial provider articles that covered the same topic (or type of news event) to ensure we compared “apples with apples”: local crime will be covered differently than the opening of a new hospital or a community event, for example.
Our primary method for classifying articles by news event and primary topic was zero-shot learning. This is a type of problem in machine learning where a model needs to predict the class (the news event or primary topic) that an article belongs to without having observed any examples of classes before. In lieu of labelled data, the auxiliary information we provided to the model was textual description: primary topics (such as “law enforcement/crime”) were augmented with definitions and other descriptive text.
For this project, we used a pre-trained MultiNLI (Natural Language Inference) sequence-pair classifier based on the open-source Bart model developed by Meta. Primary topics included law enforcement/crime, the NHS/health (including the Covid-19 pandemic), business and the economy, local government/politics, education and weather. The remaining articles could not be classified into one of the topics above and were placed into a category of “local interest stories”, such as community events or issues. Single topics were selected based on keyword weighing, with keywords highly suggestive or relevant to a primary topic given the greatest weight: for example, “sentencing” and “solicitor” are highly relevant to law enforcement/crime, the name of specific councils for local government/politics, etc. This was cross-checked against the outputs from the zero-shot classification models, which assigned a probability score for each primary topic (a news article would have separate scores for law enforcement/crime, the NHS/health, education, etc.).
Although our approach to data collection and analysis had several important limitations (in particular, we were limited to accessing the articles that News API had collected, or that news providers had optimised for search engine optimisation or posted on social media), this approach is still valuable as it significantly expands the number of articles that can be included in the analysis. In comparison, previous research had relied on manual methods to extract top articles on website homepages for content analysis.
Our much-larger sample size thus provided the most comprehensive view of the online local media landscape in the UK to date, including both the magnitude of articles published as well as their content and structure. This meant that our research significantly expands the existing evidence base, informing the broader discussion around the role of the BBC in commercial markets and the long-term sustainability of local news in an increasingly digital, post-COVID-19 world.
Read the full report here.