Why people that use search, social, and aggregators have more diverse news diets
In new research published recently, we find that people in the UK who more often use social media, search engines, and news aggregators to get news have more diverse news diets than people who mainly access news by going direct to news websites. However, these same people are also more likely to have news diets that contain a mixture of more partisan outlets on both the left and the right.
The findings, published in leading academic journal New Media & Society, are based on desktop/laptop web tracking data collected by YouGov during a one-month period spanning March/April 2017 and challenge some of the most widely held and frequently asserted claims about what online news use is like for most people. In line with previous research, they show that rather than enclosing users into filter bubbles – where algorithmic news recommenders filter out information that the user doesn’t agree with or isn’t interested in, narrowing what we know (Pariser, 2011) – social media, search engines, and aggregators are actually more likely to show people news from outlets they wouldn’t normally use, leaving them with news diets that are more balanced across multiple outlets with different editorial lines.
Why does this mean in practice? The first thing to keep in mind is that, if left to their own devices, most people do not have particularly diverse news repertoires. Many people hardly use any online news at all 1, and many others form strong habits where they voluntarily chose to go back to the same outlets over and over again. In the UK, many people repeatedly return to one outlet in particular: the BBC News website. Almost half of all news stories that people go to directly in our web tracking dataset—i.e., by navigating to a homepage and then clicking a story, or from another story on the same website—are from the BBC.
But when we look at news use on social media such as Facebook and Twitter, search engines including Google, Yahoo and Bing, and news aggregators such as Google News, the picture is very different. News use on these platforms is much more evenly distributed across different outlets. The BBC accounts for around 15 to 20% of news articles, and in some cases is behind (or is closely matched with) other popular outlets like the Guardian, the Independent, MailOnline, and the Telegraph.
This happens, we argue, because people have less direct control over news outlet selection when they use platforms. If someone decides to type ‘www.theguardian.com’ into their browser and hits ‘Enter’ they are guaranteed to be shown news from the Guardian. But if they look for news using a search engine, social media, or an aggregator they often in practice click on one of the top stories they see, and these won’t always be from a source they would have gone to directly. Due to what we have elsewhere called ‘automated serendipity’ (Fletcher & Nielsen, 2018b), probably driven in part by search engine companies deliberate efforts to create a sense of variety to keep people coming back, they will be shown articles from a range of different outlets—some of which they would not normally use. When it comes to social media, many people do not log on with the specific intention to look for news, but are ‘incidentally exposed’ (Fletcher & Nielsen, 2018a) to news while they are looking to connect with friends, entertain themselves, or pass the time. Incidental exposure is often a result of seeing articles others have shared, and because people’s networks often contain many ‘weak ties’ that connect people who would not otherwise be in contact, this can also lead to people being shown news from outlets they wouldn’t otherwise use (or shown news when otherwise they would not see any).
It’s perhaps no surprise, then, that when we look at individuals in the data – and apply scientific measures of diversity to characterise their news use – we find that the more people use search engines, social media and news aggregators, the more diverse repertoires they have (see Figure 1). Put differently, the more people use these platforms, the less reliant they are on the BBC (which plays a particularly central role in the UK), and the more balanced their news diets are over a more varied range of other newspaper, digital-born and broadcaster outlets (including the BBC).
Figure 1. Estimated diversity of people’s news repertoires by number of news accesses for each mode
It is important to stress two points. First, people who more often use search engines, social media, and aggregators do not have diets that are more skewed towards either the left or the right – their news diets are also more balanced across a range of outlets from different ends of the political spectrum. And second, many people who use platforms for news still do not have particularly diverse news diets that span the full range of what’s available online. They just have more diverse diets than people who mainly rely on direct access. Even among those who accessed at least one of the 21 news outlets we tracked, the median number of outlets used across the entire one month tracking period was just 3. This is why cross-platform studies are so useful—because they allow us to compare news use via search engines, social media and aggregators with the actually-existing alternatives, and not an unrealistic ideal.
Nonetheless, because they increase diversity, it is almost inevitable that people who more often use search engines, social media, and aggregators to get news also have news diets where more partisan outlets feature more prominently. The BBC is legally required to be duly impartial, but UK newspaper and digital-born websites operate under no such obligation – with many openly endorsing specific political parties during elections. In a relatively polarised media system like the UK (Fletcher et al., 2020), of which partisan outlets are clearly an important part, diversity necessarily means news consumption from partisan outlets.
This highlights that the use of the same platforms and services might play out differently in countries that have different pre-existing patterns of media supply and media use. And it also raises an interesting, more normative question. Is more diversity really preferable when that diversity requires the use of more partisan outlets? Although some might see exposure to more varied views as something that is broadly speaking a good thing, others might prefer a world where most people get their news from an outlet like the BBC—where, at least in the best cases, a range of different views are represented fairly. That’s a question our research can’t answer – but we hope that any subsequent attempts to do so will benefit from our work on the underlying patterns of media use.
Extending the analysis
Some readers may have a lingering question. What if those who tend to use search engines, social media, and news aggregators are very different from those who rely on direct access—meaning that the differences in diversity might not be a result of the access mode, but perhaps due to other variables like age, education, or interest in news? This is a familiar problem in social science, but not one that is easily solved.
One advantage of web tracking data is that it can reliably record people’s online media use over longer periods of time. This allows our dataset, which covers a one-month period, to be broken into four weekly parts. This essentially creates what social scientists call panel data, where the same people are measured multiple times, enabling within-person analysis as well as the between-person analysis discussed above. In other words, it becomes possible to analyse whether week on week differences in news repertoire diversity within the same person are correlated with changes in direct access, social media news access, search access, and so on (Scharkow et al., 2020). We can use the panelr package in R (Long, 2020) to build random-effects within-between (REWB) models to model within effects and between effects simultaneously. Due to space and timing constraints this analysis is not part of the published paper, and therefore has not been peer reviewed. We include it here for those who are interested.
The results in the ‘Between effects’ section of Table 1 essentially mirror the analysis described in the first part of this piece, because they refer to differences in news repertoire diversity between different people. Our focus here is on the ‘Within effects’ section, as this refers to week on week differences for the same people. They show that if someone uses direct access more than usual in a particular week, their news repertoire diversity is lower than usual (the coefficients are small because the unit is one news access, but these can compound over time). However, if they use social media, search, or news aggregators more than usual, their news repertoire diversity is higher than usual. This allows us to be more confident (but not certain) that the differences in news repertoire diversity we observe are a result of access modes rather than demographic factors or other individual-level variables that were not measured.
1 As we are interested in the diversity of people’s online news repertoires, we can only analyse the behaviour of online news users. For methodological reasons, news users are defined as anyone that accessed at least 5 news articles from any of the 21 most used news outlets during the one-month tracking period.
- Fletcher, R., Cornia, A., & Nielsen, R. K. (2020). How Polarized Are Online and Offline News Audiences? A Comparative Analysis of Twelve Countries. The International Journal of Press/Politics, 25(2), 169–165. https://doi.org/10.1177/1940161219892768
- Fletcher, R., & Nielsen, R. K. (2018a). Are People Incidentally Exposed to News on Social Media? A Comparative Analysis. New Media & Society, 20(7), 2450–2468. https://doi.org/10.1177/1461444817724170
- Fletcher, R., & Nielsen, R. K. (2018b). Automated Serendipity: The Effect of Using Search Engines on the Diversity and Balance of News Repertoires. Digital Journalism, 8(6), 976–989. https://doi.org/10.1080/21670811.2018.1502045
- Long, J. A. (2020). panelr: Regression Models and Utilities for Repeated Measures and Panel Data (R package version 0.7.3) [Computer software]. https://cran.r-project.org/package=panelr
- Pariser, E. (2011). Filter Bubbles: What the Internet is Hiding from You. Penguin.
- Scharkow, M., Mangold, F., Stier, S., & Breuer, J. (2020). How Social Network Sites and Other Online Intermediaries Increase Exposure to News. Proceedings of the National Academy of Sciences in the United States of America, 117(6), 2761–2763. https://doi.org/10.1073/pnas.1918279117