Introduction
GB News was launched in June 2021 and considers itself “United Kingdom’s News Channel” with its content independently verified as having a right-leaning bias. Despite proudly stating they are an “Ofcom regulated channel” that same government-backed regulatory body has since found them to have breached broadcasting rules 13 times with further investigations still ongoing.
Not only are they playing fast and loose with the rules they are also doing the same with their finances. At the end of last financial year, they reported a 38% increase in losses bringing the total deficit to £76m since launch. In an effort to save money, the channel has started offering voluntary redundancies, cutting 40 jobs during an election year.
While breach of due impartiality rules, financial losses and even allegations of sexual impropriety of staff members have all been reported on I wanted to gain a better understanding of the viewers it attracts. What push and pull factors are in play and what effect, if any, is the channel having on the population?
To help with this I am using two surveys, one carried out by Redfield & Wilton Strategies (15th to 16th April) and another by Johnson and Lubbock Partners (15th to 22nd April). Both are members of the British Polling Council ensuring a “standards of disclosure” which includes providing raw table data.
The Redfield & Wilton Strategies surveyed a more dedicated viewership as the respondents agreed GB News was “one of the television news channels that they would typically use for news”. Compared to only 49% of the J. C. Partners respondents who said they “currently watch” GB News resulting in the other 51% only saying they “watched it in the past” (See Table 93). Leaving the concept of “past” open to interpretation.
Push Factors
Unfortunately, the Redfield & Wilton Strategies survey whose respondents were the most dedicated was also the least comprehensive. At first, it didn’t seem much could be learned, it wasn’t surprising that a right-leaning news channel had 38% of its viewers voting for a right-leaning party. Even their 62% male-dominated viewership follows the ‘modern’ gender gap in British politics that has women being more likely to support left-wing parties than men.
One aspect of the survey that piqued my interest was the geographical distribution of viewers. Even at a high level, across 11 regions and countries of the UK, there was a distinct distribution. The question then becomes can other statistics be correlated to the same geographic distribution? Answering that question might help explain possible push factors in becoming a viewer.
When working with correlations you can use a statistical tool called the Pearson Correlation Coefficient. This measures the strength and direction of the linear relationship between two variables. This can then be converted to a probability value (p-value) which helps to determine the significance of the Pearson Correlation Coefficient. Generally speaking, a low p-value of less than 0.05 suggests that the correlation is statistically significant. It essentially translates to saying we are 95% sure the observed correlation didn’t happen by chance.
The most obvious correlation would be population size. If for each region you compare the percentage of viewers to the percentage of the population you will find a statistical significant correlation (p-value 0.01). This makes sense, as you would expect to see the viewership increase with the population.
However, this doesn’t mean that each region increases at the same rate. There was no statistical significance (p-value 0.46) when comparing the ratio of viewers to population and population percentage. This tells us something else is influencing the viewership.
The survey already found a plurality of viewers intend to vote for right-wing parties like Reform UK and 63% voted for Brexit. However, there was only a correlation with Reform UK voters (p-value 0.02) and not Leave voters (p-value 0.75). While the Brexit vote occurred back in 2016 it was also a national campaign that influenced the opinion of the general public. For example, Leave voters mainly wanted control over immigration but after the vote concerns over immigration have dropped despite there being increasing levels of immigration.
The viewers of GB News could represent a subset of Brexit voters that were attracted to Brexit due to their individual circumstances that still remain relevant after the influences of a national campaign. The same set of circumstances could be pushing voters towards a more right leaning party as right leaning voters could be part of the 69% of people that disapprove of the current government.
Right-leaning voters who disapprove of the current government will then be drawn to a more right-leaning party as a number of GB News presenters are political aligned with Reform UK, including Lee Anderson, Richard Tice, Nigel Farage and Michelle Dewberry. They could easily present a worldview that criticises the government and praises the solution offered by Reform UK.
The question remains as to what external factors are affecting the circumstances of people that make them sympathetic to these ideas. While polls had immigration as the 1st or 2nd reason for voting Leave there was no correlation with immigration changes versus the ratio of GB News viewers to population. Comparing percentage changes in census data between 2011 and 2021 showed no correlation with people being born outside the UK (p-values 0.64) or people with Non-UK passports (p-value 0.54).
The only correlation that I found that could be used as a predictor for GB New viewers was deprivation. In particular, the Multiple Index of Deprivation (IMD) where up to seven domains of deprivation are scored, including Income, Employment, Education, Health, etc. These are then weighted with different strengths and compiled into a single score before being ranked. As a result, small geographical areas within the UK that contain between 500 to 1,200 households can be ranked from least to most deprived.
Unfortunately, the IMD rank is produced independently by the devolved countries and is therefore not directly comparable. Instead of working with three different rankings, we could use the compiled score on which the final rank is based. While the methodology still differs the scores can be averaged and compared. For some reason, Scotland decided not to publish these scores which meant recalculating them based on the underlying domain scores.
To achieve this the Scottish Government released openSIMD, an open-source program that carries out the same mathematical functions using freely available statistical software. The small caveat is that the recalculated scores will not match exactly due to differences in rounding and factor analysis. Thankfully the authors of the program concluded that the differences between the official ranks were not statistically significant. After creating an average score for each region of the UK there was a clear statistical significance (p-value 0.01).
In an attempt to remove any concern over the differences in methodology, I also created a comparable UK IMD score. It was based on the work done by researchers in 2016 who created a method of adjusting the IMD score across the UK. The resulting score only relied on Income and Employment data given that both domains are already weighted heavily, between 22% and 28%, and are approximately comparable across countries. While the resulting scores from this research were made publicly available it was based on data from 2012–15. After using data from 2019–20 there was minimal change to the correlation (p-value 0.02)
This doesn’t imply that the viewers themselves suffer from deprivation, quite the opposite since 92% of viewers self-identified as working or middle class. This simply reinforces the current research that “relative deprivation was positively associated with right-wing populist votes among people with high incomes”. Deprivation can then be seen as a conduit for psychological insecurities and frustrations that erode the legitimacy of political leaders and an increase in alienation as local communities break down. Right-wing nationalism can then present itself as a necessary reaction to gain control over feelings of anxiety brought about by unwanted changes.
Pull Factors
Now that we can infer an emotional state of mind as to why people will seek out right-wing nationalism we will now investigate how the GB News takes advantage of that. To do this I will be analysing the subtitles of non-repeating shows that aired between the 8th and 22nd of April 2024. These two weeks represent the speech that occurred before and during the survey responses.
To get an accurate list of programmes during that time I retrieved an Electronic Programme Guide and removed shows that contained “Replay” in the title. The “UK National Anthem” was also ignored along with “PMQ’s live” as it is mainly a feed of parliament during prime minister questions. The final list of shows were then matched with their YouTube video counterparts which were uploaded in full. Only three shows were not available on YouTube. Finally, the subtitles of the YouTube videos were downloaded and ready for analysis.
Propaganda: Word Level
The subtitles themselves contained no punctuation which makes further analysis difficult. To resolve this a predictive model was used to apply punctuation restoration. The model itself was trained on 10 years worth of news data with a high accuracy rate. With punctuation restored, we can now use other predicative models that work at the sentence level. The first one of these models predicts the use of propaganda and can detect 18 propaganda techniques.
Instead of simply providing a score for each propaganda technique I thought it might be more useful to show the average scores over time of day. This could highlight certain presenters or shows making more use of propaganda than others. The final result does show a variation in propaganda use with the primary technique being “Loaded Language” which is when words or phrases are used to elicit strong emotional reactions.
To gain a better understanding of the types of emotions used another predicative model was applied that detects six basic emotions: anger, fear, joy, love, sadness, and surprise. As expected a similar pattern emerged but with joy and anger being the main linguistic tools used to manipulate the audience. Interestingly the morning shows have higher scores of joy which drop as the day goes on making the case that emotional language is artificially controlled rather than naturally occurring.
The next stage requires an understanding of what subjects are consistently rejected with anger or awarded by joy. To achieve this level of granularity we can gain insight into the subject matter by using a predictive Named Entity Recognition model and subjective noun extraction. Before any of that is carried out advertising slogans like “GB News” and “Britain’s News Channel” are removed to prevent them from influencing the classifying process.
The process is carried out as follows, the highest scoring emotion is given to each sentence which is then linked to the extracted entity or subjective noun in that sentence. This means for each emotion we now have a list of locations, organisations, people and subjective nouns each with their own summed score for that emotion. This data now allows us to figure out how stable or reactionary that emotion is and what concepts cause contradictions where the same concept is represented by both positive and negative emotions.
A stable emotion is one where the subject matter surrounding that emotion doesn’t change much from day to day. This was calculated by getting the daily top twenty words for each emotion. These daily top twenty lists were then merged for each emotion, removing duplicates. The results show that joy and anger were the most stable with 95 and 109 unique words appearing in the top twenty lists compared to love which contained 240 words. This suggests words associated with fear, sadness, surprise and love were used in a more reactionary way with less of an underlying theme.
For this reason, I decided to look at the total top ten words associated with anger and joy. Here we can easily see the main themes appearing and the implied contradictions. For example, themes of nationalism score high in joy with British, Britain, UK, and GB all appearing in the top 5 but Scotland and London have higher anger scores despite both being British. The same thing happens for British institutions like the police and government.
The implied contradictions in these instances can be seen as a response to defining in and out-groups. The concept of right-wing nationalism is the main driving ideology but any movement that threatens this is considered the out-group. For Scotland, this could be due to the Scottish independence movement or that social attitudes in Scotland appear more left-wing than in England. A similar left-wing bias is seen in London along with its higher levels of diversity.
Propaganda: Narrative Level
So far the analysis has been at the word level, the authors of the predictive propaganda model used earlier hoped it could “complement document-level judgments” which are inherently more complex. To give an example of this we can use one of the monologues given by a presenter at the start of their show which are then uploaded as an opinion piece to the GB News website. This particular opinion piece was about the Archbishop of Canterbury urging the Government to provide tax credits to families with more than two children.
The Archbishop’s statement in the House of Lords included research by the End Child Poverty Coalition which stated that, “removing the two-child limit would lift a quarter of a million children out of poverty.” This has wide-reaching implications including on their “educational outcomes, their mental and physical health, and their likelihood to require public support from public services later on”. With the Archbishop firmly stating that “the moral case is beyond any question.”
In this example, the leader of the Church of England, a national institution, is directly threatening the economic model of popular right-wing nationalism which is based on “native producerist deservingness” where there is “generous social policy spending on the old and families” while “non-natives, the poor, and the unemployed are stigmatized through cuts”. This speaks to the deprivation witnessed by the GB News viewers who see the poor and unemployed around them as the reason for their loss in the community. Any increase in spending towards poor families would then be seen as underserving and a threat to this model.
To remove this threat from an authoritative religious figure many things need to be done. First, all context surrounding the statement has to be removed including the research, the idea of long-term investment and the moral implications. The moral issue then has to be redirected towards protecting the nation from further debt. Finally bring into question the legitimacy of the religious figure by questioning their motives as a member of the out-group. This whole process can include contradictions, unsubstantiated and false claims for the sole purpose of propagandising the ideological subject.
For example, the first paragraph of the monologue asks why the Archbishop is getting involved in politics which is followed one paragraph latter complaining that the Archbishop didn’t previously involve himself in politics. The Archbishop is later described as a member of the “pampered, unelected, out-of-touch elites” despite the presenter himself using his unelected position to advocate for his own beliefs while he advertises himself as an “after dinner speaker and event host” who lives in Highgate, North London which has one of the most expensive streets in Britain.
The idea of a “worrying rise in crime” is brought up despite evidence to the contrary. They repeated the lie that the Church had created a “conveyor belt system of baptisms of asylum seekers” which amounted to thirteen since 2014. The Archbishop was accused of not commenting “on the huge social and moral problems facing this country” despite meeting with a youth group just ten days earlier where people engaged in “meaningful dialogue, practical exercises and reflective activities aimed at fostering empathy, understanding and resilience”.
There are many more examples of narrative manipulation but the whole point of any propaganda piece is to be consumed and not analysed. By the time you can prove the use of propaganda its effects have already taken place.
Effects on population
We now understand why the GB News viewers might be pushed towards right-wing populism and how GB News pulls them by reinforcing their ideology. Is it possible to see the effects of this on the population?
Survey Contradictions
One possible mechanism is the discovery of contradictions in surveys where there is a drop in support for a particular action based on how the question is worded. An example of this can be seen in the second survey by Johnson and Lubbock Partners. In the following survey question there is agreement amongst GB News viewers that poverty and inequality need to be tackled.
JLP Survey: Q12 — Table 65 “Tackling poverty and inequality should be the government’s top priority”
Respondents: Agree 67% / Disagree 11%
However, this is directly followed by viewers saying there is too much money going towards poor people. There is no empirical data to back up this belief so propaganda would need to be heavily relied upon.
JLP Survey: Q12 — Table 69 “There is too much reliance on welfare and benefits in Britain today”
Respondents: Agree 68% / Disagree 12%
This first question is too abstract to be tied to any political ideology allowing for a more nuanced, human view to come through. Empathy is deeply rooted in our brains and through evolutionary mechanisms was formed as a type of mind reading . Given the viewers are correlated with areas of high deprivation it makes sense that this is an issue they want to be prioritised.
However, this may simply be a result of wanting to help people seen as the ingroup since the GB News viewers are not correlated with higher levels of immigration. Empathy can be easily focused towards those deserving based on group membership and because poverty is something seen at the local level it can be seen as helping your people.
On the plus side, this might be a sign that there is less stigma towards poverty itself. Without top down governmental policies designed to stigmatise poverty, as seen in the Victorian or Thatcher era, GB News may run the risk of alienating their viewers. Even in the previous propaganda piece about the Archbishop, the presenter expressed he was “deeply concerned about growing child poverty, which must be addressed”.
To get round this GB News attacks “welfare” and “benefits”, the mechanism that supports people in poverty. This aligns with the government’s approach where a damming 2019 United Nations report into poverty summarised that the government had reduced “benefits by every means available” including “harsher penalties, depersonalization, stigmatization”. This new approach, called welfare chauvinism, allows for the existence of the welfare state but only for those deserving of its benefits.
For example, welfare needs to be strictly reserved so not to increase the national debt or be given to immigrants. To achieve this aim right-wing populist actors generate a narrative focusing on the failures of the system and creating a sense of crisis. This sense of uneasiness can be used to weaponise empathy towards the ingroup or protect the nation.
To analyse how this appears on GB News sentences that contained the word “poor” or “inequality” were compared against sentences containing the words “welfare” or “benefits”. The first obvious result is that issues relating to “welfare” or “benefits” have three times the coverage. This is despite 14.3 million people being in poverty, double the 6.2 million people on benefits. Not only does this generate a certain narrative towards people on welfare but it also protects the viewer from feelings of shame towards their nation.
Secondly, the use of emotional language is more reactionary and can’t be easily split into joy and anger. Instead, a simpler sentiment predictive model which had been trained on 124 million tweets was used. These negative and positive scores helped to show the correlation with the survey that showed people viewed both poverty and welfare as bad.
Hate Comments
The effects of this level of propaganda are most evident in the comment sections where the previous opinion piece about the Archbishop had the most upvoted comment calling him a “traitor to the UK that needs sacking and removing from EVER having a say about anything again.” Now that the Archbishop has been labelled as the out-group their right to free speech is now easily targeted.
Another area where the effects can be measured more broadly is the comment sections under each video published on YouTube. While comments can be left under the full shows they are also hidden behind a playlist tab and further into a playlist folder. Out of the 128,766 comments made over the two-week period, only 1% of comments were made under these full videos. The rest were made under shorter edited videos shown on the channel’s YouTube home page. These videos have been produced so as to be picked up by the YouTube algorithm.
To provide a broader analysis we can run those comments through a predictive model trained on 154M tweets finetuned for hate speech detection. A total of 11,136 (8.6%) comments had some form of hateful speech with gender, religion and race being the top three types of hateful comments.
To gain a better insight into which group were being specifically targeted a text classifier was used for each hateful comment. Gender, sexuality and age were simply categorised into two basic forms. The other categories were split into the top six religions and the top four ethnic groups. Obviously, further categories exist but the process is computationally complex and the current list of categories took over 48 hours to process.
Most hate was targeted at women, which might explain the lack of women’s viewership. This is followed by Muslims, men and Black / African.
If we then run that same categorising process through the subtilities of the YouTube videos seen on the channel’s home page and get the emotional values for each category we can then compare the amount of anger shown in the videos with the amount of hate seen in the comments. Unsurprisingly there was a strong statistical correlation (p-value 0.01).
How much of this online hate is flowing into the streets is unknown although not all comments will be from the UK. Despite this there is an almost statistical correlation between GB News viewers and hate crimes (p-value 0.058).
Conclusion
There are many reasons people could be attracted to right-wing populism. The simplest reasons could be down to differences in personality where people that vote for right wing populist parties were associated with lower levels of openness to experience and agreeableness. Lower levels of agreeableness and openness can then be associated with income inequality and job insecurity. For the working class viewers by living in areas of deprivation the chances of finding a job if lost or finding customers will always be a source of insecurity.
By living in higher levels of deprivation over an extended period of time, the personality of GB News viewers have made them favourable to any ideas that would improve economic stability. The trust in current politicians that promised improvements has long since eroded. Giving right-wing populists a chance to hijack these personality traits and present solutions that seem to offer stability while also carrying a right-wing ideology. For example, the reduction in immigration can be presented as a solution to the housing crisis.
This economic background won’t apply to everyone who watches GB News. A prime example of this can be seen in the upper-class GB News presenters or the leaders of Reform UK, Nigel Farage and Richard Tice, where Nigel had a Coutts bank account that required £1,000,000 to open and Richard’s previous job was a property developer and investor. Their beliefs came from a cultural understanding of what Britain should be be which is constantly under threat due to the progress of time. This is why researchers often categorise right-wing populism into economic and cultural concerns.
Given that the people with economic concerns are the larger group GB News should be focusing their attention there. For example blaming immigrants for cultural instability might not resonate with viewers who are not correlated with changes in immigration with only 39% (Table 76. Q12) said they were uncomfortable with the rate of change in their community. The latest viewing figures from BARB show a year-on-year drop of 9% meanwhile the time spent viewing increased. This might suggest a loss of viewership from the broader economic group while increasing the interest of the more dedicated cultural group.
Losing viewers from the economic group could come about once they start realising how much propaganda is needed to maintain the façade that their solutions offer economic stability. That the removal of the two-child cap will save money in the long run. That leaving the ECHR will leave us economically vulnerable and possibly worse off as we renegotiate our EU trade agreement and the Good Friday agreement. That alternative approaches to processing asylum claims will increase spending. That tackling deprivation requires a multifaceted approach involving markets, state and individuals that can be at odds with right-wing ideas. That the Reform UK manifesto costings don’t add up just like every other over promising populist party.
The biggest concern comes from the correlation between anger and an increase in online hateful speech. There is already a correlation between deprivation and an increase in hate crimes and an increase in hate crimes was associated with Brexit. This is especially worrying since the rhetoric of the Brexit campaign will be similar to what will be broadcast on GB News. Finally the targeted anger towards certain out-groups could easily normalise feelings of hate. Taken together with personalities that have reduced openness to others and you have a fine tuned hate factory.
