In our Is Britain Pulling Apart? project we’re interested in social interactions between different types of social groups. One characteristic I’m particularly interested in is social connections by newspaper choice. Within the UK, newspapers can reveal elements of both our social status and social attitudes. Earlier this week, for instance, I noticed this tweet which juxtaposed attitudes towards welfare claimants between two popular papers.
This difference in editorial position is what makes social connections by newspaper readership interesting. If people are only reading the Daily Express/Daily Mail positions about people on benefits they might harness a different perspective to someone reading a paper such as the Guardian. If people are surrounded by people who read very different papers to themselves, they are perhaps exposing themselves to a variety of positions, attitudes and statistical evidence. Newspapers might aim to tell the truth, but are selective in which truths they tell.
Looking at the demographics of newspaper readers shows they are attracting different markets. The graph below shows the proportion of readers who are graduates (commonly associated with the broadsheet/tabloid distinction) on the vertical scale, whilst the horizontal scale shows the proportion voting Conservative, with the Guardian and Mirror to the left and Mail and Telegraph towards the right. Many readers of this blog might broadly have guessed which paper was which even if they weren’t labelled. Such distinctions aren’t as obvious amongst television news; indeed, those identifying as mostly receiving news from a particular channel are much more similar in their demographics than those reading specific papers.
Demographics of newspaper readers (UKHLS, wave 3)
Characteristics of viewers of TV news (UKHLS, wave 3)
Within our current project we’ve presented several papers which have looked at the social mixing between consumers of different types of newspapers (see our website for slides). Homophily, the effect of birds of a feather flocking together, is central to our analysis and we are interested in whether people tend to associate with those reading similar, or different, newspapers to themselves. This is a non-trivial question, given the different perspectives and viewpoints associated with papers. However, due to the availability of data, this has usually been within-household social connections by newspaper readership. Commonly, one of the questions we’ve received has asked whether studying within-household newspaper is appropriate as often the household will read the same paper. In this blogpost I’ll outline why using a household survey, such as Understanding Society (UKHLS) is appropriate for newspaper connections.
A few practical notes should be made. Firstly, that this blogpost ignores households consisting of just one person, given there can be no sharing in those homes. Secondly, this analysis uses UKHLS data, which asks how people most commonly obtain news and, if they state via newspapers, which paper they most commonly read. Therefore, this discussion is only of the 55% of nearly 50,000 respondents who named a newspaper. It is likely some people read multiple daily papers, and also some whom get most of their news from the internet or television are also newspaper readers but such people are excluded from this analysis.
Our presentations often receive a question about whether all residents within a household read the same paper due to physical access to it. It is easy to think of circumstances where this will be true, especially for the bulkier Sunday papers. However, there are also many situations where people living in the same home might not have access to the same paper, for instance, through one resident reading a colleague’s paper during their lunchtime at work, or through leaving the paper on the train at the end of the commute. Similarly, there might be circumstances where parties have little interest in papers within their home – perhaps a father who reads the Financial Times and a son who reads the Daily Star will have little interest in each other’s newspapers.
Across all households with two or more respondents (UKHLS, wave 3), 91% live with at least one other who shares their newspaper status, if no affiliation in included. However, when we look solely at those who read a newspaper, 48% live with someone who reads the same paper. The highest levels of homophily are for the Herald (63%), Telegraph (56%), and Daily Record (55%). The lowest is for the Financial Times (16%). Identifying as reading a specific newspaper doesn’t necessarily mean you live with others with the same identity.
Newspaper readers living with someone who also reads same paper (UKHLS, wave 3)
Of course, it’s plausible this lack of homophily is caused by people not identifying as a reader of any newspaper. This is often, but not always, the case. In 14% of two-person households the residents read different newspapers. At least 2 different papers are read in 29% of three-person households, 35% of four person households and 48% of homes with five or more people.
Papers per household (UKHLS, wave 3)
Similarly, it’s not necessarily the case that couples read the same newspaper, with the differences occurring outside of partners. In around one-third of cases where couples both identify as newspaper readers different papers are selected. A simple logistic regression can show this trend, using those in relationships from UKHLS wave 3. The first model predicts readership of newspaper based on age, gender, education (three categories) and occupation (microclass). The pseudo-R2 (basically, proportion of variance explained) ranges from just .021 for the Mirror to .14 for the Sun. The second model adds a variable of whether the partner reads that paper. This always provides a notable improvement in the models, but generally an improvement of between .14 (Mirror) and .23 (Telegraph), aside from a much smaller effect for the Financial Times (.06). Finally, adding information about type of paper the partner reads (broadsheet, tabloid, Scottish, regional, or ‘other’) provides a further small increase. Clearly knowing what paper someone’s partner reads improves prediction of what paper they will read, but it only explains a small part of the story.
Pseduo-r2 of newspaper read (UKHLS, wave 3)
Therefore, it appears there is a justification for analysing newspaper readership using within-household social connections. We have mapped this in various ways, using social network analysis, log-linear modelling and multiple correspondence analysis in our presentations thus far. Whilst we are happy that within-household patterns can be mapped, there are several other factors which need to be controlled for. Social network analysis provides an interesting framework for looking at combinations of newspaper readers and demonstrations some of these issues. The first is the distinction between newspaper patterns in certain areas. For instance, the two sociograms below show the patterns of overly connected newspapers for England (top) and Scotland (bottom).
Network of over-represented within-household newspapers in England (UKHLS, wave 3)
Network of over-represented newspaper combinations in Scotland (UKHLS, wave 3)
Newspaper patterns are different generally in England and Scotland, as evidenced by the large number of Scottish newspapers such as the Daily Record, the Herald and the Scotsman and decreased readership of English-based papers. Within England the Daily Mail (a right of centre paper) is most popular whilst in Scotland the Daily Record (a left of centre paper) is the most read. Similarly, evening and regional papers in England appear to be read in households which also get right-wing or News UK papers, whereas in Scotland there are more common in households reading the Guardian than the Times. The Guardian is also more strongly connected to other broadsheets in England than in Scotland. Therefore, analysing readership connections needs to take into account not only are different papers available in England and Scotland, but also that patterns of homogamy by readership differs also.
Similarly, there are age-cohort effects associated with consumption patterns. The two sociograms below show the networks for younger couples (both partners under 50, top) and older couples (males over 50 and females over 45, bottom). The greater number of ties for the younger cohort signifies a great diffusion of multiple newspapers between partners.
Network of over-represented newspaper combinations for younger Scots (UKHLS, wave 3)
Network of over-represented newspaper combinations for older Scots (UKHLS, wave 3)
For the older cohort, patterns identifying with different papers commonly refers to the Daily Record and one other, or an evening/regional paper and one other (aside from the Herald/Times connection). Removing the Daily Record would cause the composition of the network to break down. For the younger cohort, removing the Daily Record would maintain a well-connected network, with the Sun, Mail and an evening paper at its centre. Thus, whilst an alternative viewpoint for older Scots usually includes receiving the Daily Record’s editorial position, for younger Scots there is a greater blend of attitudes and opinions being addressed. Therefore, incorporating age-cohort effects is also important.
An area I was concerned with when first using the newspaper variable in the UKHLS was the possibility of interviewer bias. There is an argument the US General Social Survey has suffered from interviewers being inconsistent in gathering the number of friends individuals hold, and there is potential for the same to occur with asking people about newspapers of choice. However, there is no evidence that some interviews were more, or less, likely to obtain valid newspaper data.
Interviewer effects for valid newspaper data (UKHLS, wave 3)
There appears to be no empirical rationale why studying within-household newspaper patterns should be problematic. Whilst instinctively it might sound like those who live together have physical proximity to the same papers, there is little evidence that people who live together identify themselves together. There are many cases where people, and partners, living together identify with different papers and mapping these, using a variety of methods, can identify important social trends. Given the widely differing social outlooks portrayed by newspapers, studying homogamy and homophily patterns enable us to see whether people’s social connections exposes themselves to people who share their values and attitudes or provides diversity of opinions and arguments. Whilst across-household data might preferable, this is difficult to obtain and existing social surveys provide within-household information which enables us to see such trends.