The Curious Case of Candidate Sanders

27+

This is the sixth installment of Epsilon Theory’s Election Index. Our aim with the feature is to lay as bare as possible the popular narratives governing the US elections in 2020. That includes narratives concerning policy proposals and candidates found in the news, opinion and feature content produced by national, local and smaller outlets.

Our first edition covered April 2019, and included detailed explanations of each of the metrics we highlight below. If this is your first exposure to our narrative maps, analysis or metrics, we recommend that you start with that primer.


Almost every successful political campaign is built around a missionary structure which must promote two narratives simultaneously. Two very different narratives. The first is that the campaign’s chosen candidate is winning – or at the very least accelerating. The second is that the candidate is getting an unfair shake, especially in the media.

Trump’s supporters didn’t just believe their candidate couldn’t get fair treatment in the media in 2016 – they made their contention a core plank of the platform. In 2020, it has been a frequent claim from the Biden camp, too. Supporters of Warren, Buttigieg, and Klobuchar have all claimed this at one time or another. Harris, too. I’m honestly not sure there ever were any Steyer supporters, but I’m sure they would have said the same. For the #YangGang, celebrating media martyrdom has become a pastime. I mean, consider that a briefly popular meme sprang up in which network indifference to Yang was parodied by replacing Yang’s image on debate lineups with headshots of various other famous Asian-Americans. Everybody whines about who is being favored in the media.

So, Bernie Bros, welcome to the club.

But is it true?

Beginning in late 2018, instead of predicting this kind of thing, we decided to start observing it. Measuring it. We measured the sentiment of language used in news about each candidate. More importantly, we think, we measured the consistency of language used in news about each candidate, both internally and with the news about the election in general. In other words, we sought to measure the common knowledge about the election and each of the candidates.

Based on that analysis, we have been convinced that Joe Biden would underperform expectations and that Bernie Sanders would outperform. We think this has largely proven true. But it is also true that the narrative landscape has shifted somewhat.

Now that we are on the doorstep of the Iowa Caucus, here are four things we think about primary election narratives:

  1. That the 2020 election narrative is still about identity and an unequal America;
  2. That Bernie Sanders is the candidate with the most cohesive narrative, which happens to be in sync with that global election narrative of identity and inequality;
  3. That media coverage of Sanders began a concerted negative and off-narrative drift in December 2019 and January 2020; and
  4. That certain key publications appear to bear a disproportionate (and curious) share of this divergence in sentiment and a previously cohesive narrative.

Now let us present why we think each of those things.


The Meta-Narrative: “The 2020 Election is About Identity and Inequality

First things first, though. Throughout this note, we will refer repeatedly to three dimensions of narrative structure we track: sentiment, cohesion and attention. Sentiment is simple. We use a standardized dictionary to track positive and negative words across articles. Nothing magical. Nothing special. Profoundly flawed and prone to misinterpretation when used on its own, so take care. Cohesion is our measure of the internal similarity of language of all language across a single topic. Think: Is everyone who writes about this thing using the same words and phrases? Attention is our measure of the external similarity of language of one topic within another topic. Think: how similar is the language used to write about sign-stealing to language used in all coverage of the Houston Astros? The intention is to replace the much cruder concept of “volume” of coverage with a concept that captures the directional flow of language, phraseology, taxonomy, analogies and other narrative-supportive and narrative-adjacent ideas.

Got it?

We turned our attention measure to the question of election topics once again for the most recent period. And once again, we came away with the same conclusion: the language being used in news articles, blogs, op-eds and feature pieces across topics is most closely aligned with discussions of identity and inequality. No other topic is remotely close. Impeachment, the Economy, National Defense and War – even Healthcare are remote also-rans in terms of how the media incorporates related language as frames for all election news.

In other words, what we are saying is that the language of identity and inequality connect stories that are not nominally about that at all. Its critical language permeates immigration, healthcare, economic and foreign policy discussions alike. It permeates debate coverage, opinion pieces, news pieces and human interest features.

Because this is an unusual way to look at media coverage, some find it is useful to visualize it.

Below you will find a series of network graphs of articles published about the election during our measurement period using software from our friends at Quid. In each, the bold-faced, highlighted nodes in the graph represent articles about a different topic. A higher attention measure in our parlance will usually correspond to topics which have stronger concentrations at the center of a network, with more language shared with other related (and sometimes unrelated) topics and articles. We think these reflect common knowledge about topics which are universally accepted and relevant within media.

Election Coverage Attention Map – Demographic Identity

Source: Epsilon Theory, Quid

Election Coverage Attention Map – Income and Wealth Inequality / Class Identity

Source: Epsilon Theory, Quid

Election Coverage Attention Map – Impeachment

Source: Epsilon Theory, Quid

Election Coverage Attention Map – Borders and Immigration

Source: Epsilon Theory, Quid

Election Coverage Attention Map – The Economy and Trade

Source: Epsilon Theory, Quid

Election Coverage Attention Map – National Security and War

Source: Epsilon Theory, Quid

Election Coverage Attention Map – Health Care and Health Insurance

Source: Epsilon Theory, Quid

Bernie Sanders is the On-Narrative Candidate

As we have recounted in previous installments, Bernie Sanders has consistently had his story told more clearly in the media than any other candidate. What we mean is that these stories used the most similar and most interconnected language. Yang and Bloomberg, on the other hand, have struggled mightily to produce a clear narrative in the minds of US political media. Other than describing Bloomberg as a billionaire and talking about Yang’s UBI plans, articles about them are all over the map.

We think it is almost self-evident that Sanders is the candidate most attached to the chief ‘meta-narratives’ of the election described above, but this is also quantifiable. The chart below is similar to the topical attention measure described in the first section. In short, this is how much more or less than the average attention articles about each candidate have within the broader universe of election coverage. Think of it as a measure of ‘linguistic correlation’ between the articles written about a candidate and all articles written about the election.

Note that Yang and Bloomberg are missing from this chart. They simply don’t have enough articles for the full period to make them an apples-to-apples comparison; if included, their values would both be well below Warren’s, and would not at all change the observation about Bernie’s unique connection to the framing of election coverage.


The Internal Narrative Break on Sanders

If you ask a Sanders supporter, my suspicion is that they would concede very little of the above, but would tell you that they feel like the campaign has come under major assault in media in the last 2-3 months. So, are they right?

Well, in the last two months, the internal cohesion of language used in Sanders coverage has cratered. Its relationship to broader election narratives has collapsed. And the sentiment of articles – relative to other candidates and on a standalone basis – has become sharply more negative. Those are the major dimensions of narrative structure we measure. All of them have deteriorated quickly in December and January.

Our attention measure for Sanders – the linguistic relationship of his coverage to broader election narratives – was actually the lowest of all candidates over this two month stretch.

Source: Epsilon Theory, Quid

Narrative cohesion measures were no kinder to Sanders. After months of on-narrative treatment from everyone writing about Bernie Sanders and his campaign, readers finally started to see a sharp divergence in the language being used.

Source: Epsilon Theory, Quid

The sentiment of what was being written about Sanders also became very negative during this period. The units on the graph below are raw sentiment units on a scale from -5 to 5, which is the theoretical range an article might score. In practice, a score of around 1.5 for any article reflects very positive language, and -1.5 or so would generally reflect very negative language. On an average basis across dozens of articles, anything approaching those levels would be extreme. It doesn’t take much sleuthing to suppose that a major cause of the “negative” sentiment was related at least in part to coverage of the “liar” spat with Elizabeth Warren, although in the latter’s case, it did not come at the expense of a previously high attention narrative.

Source: Epsilon Theory, Quid

The Sources of the Sanders Narrative Break

To some extent, news, opinion pieces and feature pieces about Sanders became more negative because the actual news event being covered was more negative. When a candidate is involved in very public spats in which both participants call their opponents “liars”, any systematically applied sentiment dictionary is going to call it “negative.” Comparisons of the trajectory in sentiment among major outlets leave that explanation looking a bit threadbare, however.

Below we compare eight of the most shared and most viewed web-based outlets for Sanders news and general election news. The below chart shows, for each outlet and for each month between January 2019 and January 2020, the percentage of articles quantified by Quid’s default news-scoring dictionary as “negative.” Each vertical bar represents a progressive month over this period.

Source: Epsilon Theory, Quid

There are two takeaways: first, yes, every outlet appears to have generally increased the extent to which they use language with negative affect to cover the Sanders campaign. For the reasons described above, that shouldn’t be taken as a sign of “bias” per se. But the second takeaway is concerning: four of these key outlets – the New York Times, Washington Post, Reuters and Huffington Post – used dramatically more negative language in their news, feature and opinion coverage of the Sanders campaign in the month of January 2020.

We are always skeptical of relying on sentiment scoring alone; accordingly, we also examined which outlets drove the breakdown in the previously cohesive use of language to describe Bernie Sanders, his policies and his campaign in the media. In other words, which outlets have “gone rogue” from the prevailing Sanders narrative? Are they the outlets who chose to stay “neutral” or at least relatively less negative in December and January? Or can we pin this on the ones who have found a new negative streak in their Bernie coverage? Is there even a relationship between the rapid shift in sentiment by some outlets and the breakdown in narrative structure?

Oh yeah.

The below chart shows the difference in the attention of each outlet’s coverage within ALL Bernie Sanders coverage for December 2019 and January 2020. If an outlet’s value is +10%, that means that they were much more “on-narrative” than average with their Sanders coverage. If an outlet’s value is -10%, they were much more “off-narrative.”

What do these two charts together tell us?

I think they tell us that the Washington Post and, to a lesser extent, the New York Times experienced a shift in the nature of their coverage, the articles and topics which they included in their mix, and the specific language they used in the months of December and January.

I think they tell us that change was unusual in both magnitude and direction (i.e. sentiment) relative to other major outlets. Their coverage diverged from the pack in language and content.

I think that change was big enough to create the general breakdown in the Sanders that observers have intuitively ‘felt’ when they consume news.


How should we respond to things like this?

In some ways, you could make the argument that the shift is a bit of a normalizing influence. After all, we have argued for the better part of a year – actually, we did it again in our first two bullets above – that Senator Sanders appeared to be favored in the way media covered him and his campaign, framing nearly every issue in terms of his preferred language and policy stances. For that reason, we have prescribed caution in reading news coverage of his campaign (and Biden’s, whose campaign has been universally loathed by media outlets of all colors).

Even if it’s for a different reason, we still prescribe that caution.

But the sharp, nearly instantaneous move and divergence is equally worrisome. Why now? Should we be concerned that a publication which used its editorial page to endorse two candidates suddenly experienced a simultaneous change in tenor of its news coverage?

Not a trick question. Obviously, the answer is yes.

It doesn’t mean there is intentional bias being injected. It doesn’t mean that bad people are doing bad things. Data can’t tell you that, no matter how much anyone tries to argue that it can. But in our view, it IS enough to cast an even more suspicious eye than usual toward the election news we will receive from certain outlets in the coming weeks.

Clear eyes, full hearts.

27+

5
Leave a Reply

Please Login as a Paid Member as a Paid Member to comment
  Subscribe  
newest oldest
Notify of
Nicholas Allen
Member
Nicholas Allen

Is it possible to incorporate node weight by view-count into the quid visualizations? It would be interesting to see how the weighting shifts based on cohesion etc.

Also, I’m still not sure I understand the ‘Attention’ dimension. In here you explain it first as:

“external similarity of language of one topic within another topic. Think: how similar is the language used to write about sign-stealing to language used in all coverage of the Houston Astros?”

…and later as:

“the linguistic relationship of [Sanders’] coverage to broader election narratives”

As far as I can tell you use ‘attention’ to refer to the cohesion of how off-topic articles refer to the topic in question. Is that correct? If so, it seems like ‘attention’ is a bit of a misnomer; instead of ‘cohesion’ and ‘attention’ I’ll probably substitute ‘internal cohesion’ and ‘external cohesion’ or somesuch. I still expect I don’t understand the ‘attention’ metric though.

3+
tany
Member
tany

It sounds like one could say that Sanders has experience and increase in “bad press” in December & January?

These last two months his campaign has been performing better in the polls, moving from second to close second and finally into first place in Iowa (https://www.realclearpolitics.com/epolls/2020/president/ia/iowa_democratic_presidential_caucus-6731.html) while generally doing better at the national level as well (https://www.realclearpolitics.com/epolls/2020/president/us/2020_democratic_presidential_nomination-6730.html).

Could this turn in attention be in response to Sander’s performance in the polls?

0

The Latest From Epsilon Theory

Join Us. No, Really.

By Rusty Guinn | February 25, 2020

The Fall of Wuhan

By Ben Hunt | February 24, 2020

Kitchen Sink It

By Rusty Guinn | February 20, 2020

Thanksgiving

By Rusty Guinn | February 19, 2020

ET Live! – 2.18.2020

By Rusty Guinn | February 18, 2020

Love in the Time of COVID-19

By Ben Hunt | February 17, 2020

The Industrially Necessary Doctor Tedros

By Ben Hunt | February 16, 2020

Options

By Rusty Guinn | February 12, 2020

Body Count

By Ben Hunt | February 10, 2020

We Hanged Our Harps Upon the Willows

By Rusty Guinn | February 10, 2020
DISCLOSURES

This commentary is being provided to you as general information only and should not be taken as investment advice. The opinions expressed in these materials represent the personal views of the author(s). It is not investment research or a research recommendation, as it does not constitute substantive research or analysis. Any action that you take as a result of information contained in this document is ultimately your responsibility. Epsilon Theory will not accept liability for any loss or damage, including without limitation to any loss of profit, which may arise directly or indirectly from use of or reliance on such information. Consult your investment advisor before making any investment decisions. It must be noted, that no one can accurately predict the future of the market with certainty or guarantee future investment performance. Past performance is not a guarantee of future results.

Statements in this communication are forward-looking statements. The forward-looking statements and other views expressed herein are as of the date of this publication. Actual future results or occurrences may differ significantly from those anticipated in any forward-looking statements, and there is no guarantee that any predictions will come to pass. The views expressed herein are subject to change at any time, due to numerous market and other factors. Epsilon Theory disclaims any obligation to update publicly or revise any forward-looking statements or views expressed herein. This information is neither an offer to sell nor a solicitation of any offer to buy any securities. This commentary has been prepared without regard to the individual financial circumstances and objectives of persons who receive it. Epsilon Theory recommends that investors independently evaluate particular investments and strategies, and encourages investors to seek the advice of a financial advisor. The appropriateness of a particular investment or strategy will depend on an investor’s individual circumstances and objectives.