ET Election Index: July 31, 2019

This is the fourth 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 goal is to make you a better, more informed consumer of political news by showing you indicators that the news you are reading may be affected by (1) adherence to narratives and other abstractions, (2) the association/conflation of topics and (3) the presence of opinions. Our goal is to help you – as much as it is possible to do – to cut through the intentional or unintentional ways in which media outlets guide you how to think about various issues, an activity we call Fiat News.

Our goal is to help you make up your own damn mind.

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.


Election Narrative Structure as of July 31, 2019

Source: Quid, Epsilon Theory

Commentary on Election Narrative Structure

  • We officially think there is a 2020 election narrative.
  • The common knowledge is that the 2020 election is a referendum on race, gender and identity.
    • This doesn’t mean we agree or disagree with this characterization.
    • This means that this is what everyone thinks everyone thinks the election is about, at least as promulgated by US political media.
  • Every highly connected cluster in the narrative structure from the month of July is charged with and defined by this language.
  • Asylum seekers and immigrants, the black vote, the narrative of electability surrounding women and gay candidates, and ‘the white vote from the rust belt’ loom large in the center of and in connections between nearly all 2020 election coverage.
  • Sentiment in coverage has also started to crystallize in a more dramatic way:
    • Sen. Harris and Biden have taken the raw end of this exchange, and in a more coherent, higher attention way than before.
    • In contrast, Sanders and Warren have received glowingly positive language in their media coverage.
  • We also note that Trump himself has begun to insert his presence into the narrative structure, despite being less present on the formal campaign trail.

Candidate Cohesion Summary

Commentary on Candidate Cohesion

  • Post-debate Sen. Harris has a much more coherent narrative structure than in prior months. Unfortunately – as noted shortly – it is one loaded with negative language, especially relating to Harris’s law enforcement background and spars with former VP Biden.
  • Biden’s coverage has been similar to Harris’s: more coherent, but coherent in its skepticism that he is a candidate that can win, skepticism that his record is sufficiently progressive to energize the Democrat base, and skepticism that he will address the race, gender and other identity issues lying at the center of the 2020 election zeitgeist.
  • Sen. Warren is a bit of an enigma. In many ways, her narrative strikes us as a “poor man’s Sanders” – less internally cohesive, less in tune with the zeitgeist, and positive…but not quite as positive as Sanders. But qualitatively, she is increasingly entangled with the same anti-corporate power, anti-inequality base and narratives that are most strongly associated with Sen. Sanders.
  • As per usual, media accounts of Gabbard and Yang are indifferent, varied and largely presented in context of other candidates. After the shock of a surprisingly positive performance in initial debates, Buttigieg content has reverted back to prior incoherent mixtures of general “round-up” content and narrow issue pieces.
  • The media seems to regard O’Rourke with a collective “meh”. They know who he is, and they’ll cover him, but the days of magazine covers and strong common knowledge about what “Beto means” appear to be gone for the time being.

Candidate Sentiment Summary

Commentary on Candidate Sentiment

  • Sens. Warren and Sanders – perhaps unsurprisingly, given July’s emphasis on health care – were head and shoulders above the rest of the candidates in terms of coverage sentiment.
  • This is standard fare for Sanders at this point, but only a June/July development for Warren, who appears to have attracted meaningfully more positive language from political media accounts.
  • Yang and Buttigieg were the only other candidates whose language we would regard as positive.
  • Gabbard, Biden and Booker have cemented their place in the cellar. Media accounts of their candidacies are routinely negative, emphasize electability concerns, highlight conflicts/spats with other candidates, and bring out claims of hypocrisy.
  • For this reason, we would be very cautious in our consumption of Gabbard, Biden, Sanders and Warren news, where we think that emerging narratives have made it more likely that ‘news’ content will be infected with affect and affected framing, whether intentionally or unintentionally.

Candidate Attention Summary

Commentary on Candidate Attention

  • As noted before, Harris is very much in line with the July election Zeitgeist, but we regard this as a function of negative coverage. We think that undecided voters should tread carefully when consuming and reading ‘news’ about Sen. Harris, whose jabs at Biden were quickly transformed into claims of hypocrisy, assertions of a weak position to argue on issues of inequality (i.e. “Kamala was a cop!” narratives), and unelectability concerns.
  • Buttigieg has faded from connection to the language used about the election as quickly as he rose, which is not uncommon for strong debate performers who were previously minor candidates.
  • It is Beto whose disconnection to the zeitgeist has been more striking.
  • We note that Warren’s attention scores remain low, despite positive sentiment and cohesion. We think (this is our judgment / opinion, not something present in the data) that this is a function of two things:
    • Many of the positions Warren is associated with, Sanders is more associated with. In coverage, this means that Sanders tends to get the lion’s share of relationship to these key electoral issues.
    • Warren’s status as a policy wonk has meant that she has focused less on the race, gender and identity issues that we argue represent the 2020 election zeitgeist.
  • For better or worse, if Warren were to refocus efforts on participating more actively in the identity-related narratives that we believe represent the common knowledge about what the 2020 Election “is about”, we think she would emerge further as a leading candidate.
  • In the meantime and absent that change, based on our views about the influence of media-driven common knowledge effects, we think that among major candidates, Sanders will outperform most expectations, and that Biden will continue to converge to his more negative narrative.
  • This also means these are the candidates where we would be most cautious that media sources might be influencing how they want us to think about the news pertaining to them.

Election Index: Beyond the Debates – 6.30.2019


This is the third 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 goal is to make you a better, more informed consumer of political news by showing you indicators that the news you are reading may be affected by (1) adherence to narratives and other abstractions, (2) the association/conflation of topics and (3) the presence of opinions. Our goal is to help you – as much as it is possible to do – to cut through the intentional or unintentional ways in which media outlets guide you how to think about various issues, an activity we call Fiat News.

Our goal is to help you make up your own damn mind.

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.


Election Narrative Structure as of June 30, 2019

Source: Quid, Epsilon Theory

Commentary on Election Narrative Structure

  • Our previously expressed views – that adoption of active narratives into public common knowledge would cause Biden to fade in favor of Sanders, Warren and Harris – proved largely accurate (although we did not anticipate how meaningful the bump to Harris would be).
  • Our data covers the whole month (not just the post-debate period), but it is striking that the debate-specific coverage in the southeast quadrant of the graph is so distinct and separate from the rest of election coverage. Some of this is just a reflection of pre-and-post-debate narratives changing, but we think this distinction is indicative of very strong narratives in media that are being promoted by outlets in spite or or in response to debate outcomes.
  • The most connected language and most central articles, however, are not candidate- or issue-specific. They are identity-related, articles about “White America”, ‘Black Voters”, ‘Hatred, Prejudice and Rage”, and the candidates capable of representing and achieving electoral success by appealing to each. In the wake of the Harris/Biden busing debate issue, the media have taken this lens to a new extreme.
  • President Trump has entered the narrative fray on 2020, and coverage of his comments and campaign launch represented a meaningful change in the overall network structure.
  • The debate has not reshaped the narrative of “Issues That Matter”. That continues to be built around democratic socialism and more general leftward inertia. Both fear-driven articles and those with a favorable temperament toward these policies and the inequality faced by the poor, students and migrants are far ahead of competing economic, abortion, impeachment, foreign policy or entitlement program policies.
  • At a high level, we think these are the issues that media outlets are promoting as “what matters” in 2020:
    • Social equality (race, sexuality, gender and treatment of immigrations/asylum-seekers)
    • Economic equality (income and wealth)
    • Healthcare
    • Education, Student Debt
  • We think these are the issues that are otherwise absent so far from any media-promoted narrative about the 2020 election:
    • Trade
    • Foreign policy / military
    • Abortion / reproductive rights
    • Climate change and the environment (surprisingly)
  • As we gauge our own sensitivity to and focus on certain issues, we would be conscious of Fiat News nudges to “care more” about certain issues than others, and to be sensitive to those cases where certain positions are more or less attached to various candidates.

Candidate Cohesion Summary

Commentary on Candidate Cohesion

  • Despite a generally panned debate performance, Sanders remains at the top of our cohesion score, both for the primary process to-date and for June specifically. That means that the language used about Sanders continues to be internally consistent. The Sanders narrative is common knowledge.
  • A Buttigieg narrative, however, has emerged following a debate performance that crystallized in a substantially higher rate of pieces characterizing him as erudite, well-prepared and intelligent, in addition to an almost obligatory reference to the recent police-involved shooting event in South Bend.
  • While Biden content to-date has referenced a generally consistent – if negative – internal narrative, the internal cohesion of Biden stories cratered in June. We think much of this pertains to the conflict between pre-debate stories about Biden as the candidate with the greatest potential among black voters, and post-debate stories positioning Biden based on a segregationist-friendly past. It further muddies the waters of an already challenging narrative to manage.
  • The Sen. Harris narrative was less cohesive in June – a result, we think, of broader, further-reaching coverage after she emerged as a more prominent candidate following her first debate performance.
  • We would continue to be cautious in our consumption of Fiat News about Sanders, where we expect coverage to continue to be auto-tuned to existing narratives of what Sanders is as a candidate. We would likewise be cautious of Fiat News in news about Buttigieg and Biden, where significant change in narrative cohesion indicates to us potential new missionary activity and attempts to change that narrative.

Candidate Sentiment Summary

Commentary on Candidate Sentiment

  • The biggest positive movers in sentiment of media coverage in June were Sanders, Warren and Buttigieg, all of whom were already relatively positive to begin with. But social equality, economic equality and socialist-friendly narratives were especially powerful in June coverage.
  • The negative sentiment attached to Biden narratives persisted in June, and worsened slightly after additional negative coverage of his debate performance and poll results.
  • We think that sentiment, as manifested through Fiat News and affected language, is perhaps the easiest narrative structural element to spot. It is also, however, very easy to slip into confirmation of our preconceptions. When sentiment aligns with our predispositions, we tend to see it as reasonable, factual language. When it is misaligned, we tend to see it very clearly as biased.
  • For this reason, we continue to recommend caution in all news and analysis judging Biden, Sanders, Warren and Buttigieg, regardless of whether the sentiment strikes us as ‘reasonable’ or ‘justified’ given recent events.

Candidate Attention Summary

Commentary on Candidate Attention

  • If you are surprised that Harris has not risen in terms of attention in June, don’t be. Our attention measure is not a measure of how much coverage a candidate or topic receives, and while Buttigieg and Harris were the big winners from June on that dimension, our measure is one of the consistency of candidate/topic language with the overall narrative. Despite a powerful group of post-debate articles arguing for Harris as the ‘winner’ of the first round of debates, the language used to describe Harris’s candidacy remains disparate and distinct from most of the language used to describe ‘what matters’ about the election.
  • In other words, we think that Harris gained in polls and favorability through what was perceived as a rousing personal defeat of Biden, the front-runner.
  • We think that this victory hasn’t really resonated in narrative world, because it isn’t connected to the common knowledge about what the 2020 election “is about”. We think that means that – barring a change in those narratives – observers expecting Harris to demonstrate continued momentum against Warren and Sanders may be disappointed.
  • Sanders is the candidate whose own narrative is most ‘on-narrative’ with the election more broadly. In other words, what Sanders is about (in media) continues to be what the 2020 Election is about (again, in media).
  • The biggest positive mover, as elsewhere, has been Buttigieg. We think that the rise in attention here is more sustainable, given that Buttigieg’s message, sentiment and cohesion appear to have been adopted by media outlets fairly readily and quickly.
  • That also means that we would expect to see more coverage of Mayor Pete’s candidacy that seek to tell a story through Fiat News. We would remain mindful of this in our news consumption.

ET Election Index: The First Debate, Part 2

This content is related to the Epsilon Theory Election Index, a series we introduced here in hopes of better informing citizens and voters about the political narratives present in US national media.


With both cattle call Democratic debates in the books, what’s the narrative? What does everyone know that everyone knows? Below, as before, we examine our standard network graph, built on linguistic similarity to identify the internal consistency and attention on various topics and candidates

Source: Quid, Epsilon Theory

Our takeaways from this coverage:

  • The narrative from the second part of the first debate boils down to two stories told frequently, with intensely linked language:
    • That the debates and the election are about identity narratives, the credibility / authenticity attached to them, and the ongoing negotiation of each candidate’s cartoon.
    • That Kamala Harris dealt Joe Biden a body blow.
  • Nearly the entire northeast quadrant of the graph is populated by social questions and identity: Buttigieg’s comments about what a Christian is and questions about his handling of the shooting of a black citizen by a white police officer in his town, Swalwell’s repeated references to generations and the passing of a torch, and above all, Harris’s jabs at Biden over bussing and his past association with segregationists.
  • As a related issue, we also note that the language (not topics, but affect) driving much of the clustering appears to be that of the verbal jabs and attacks delivered. Clusters typified by issue and policy response language (mostly on the far south axis of the visualization) are almost completely sundered and isolated from these much higher attention, more cohesive identity clusters.
  • The Biden narrative sentiment post-debate has been as negative as we have noted in our pre-debate narrative analyses.
  • In essentials, a lot of media wanted to talk about Harris and her jabs at Biden and a similar number wanted to talk about questions of identity and social fairness/equality. A very small number of articles, on the other hand, were devoted to discussions of policy. The overlapping language between those universes of articles was shrinkingly small.
  • What has the narrative not been?
    • Donald Trump, perhaps surprisingly. His tweets, reactions, and even any of the common linguistic references to Trump were almost completely divorced from the core content and narratives covering the debates.
    • Climate Change. So far, the stories being told about the debate-side of the election have been almost silent on this topic, although in fairness, it was not a major topic for questions.
    • Bernie Sanders. While front-and-center, with plenty of air time and a lot of prominent remarks, Sanders has captured almost none of the post-debate narrative. This is quite a turn from pre-debate media, which consistently published its most cohesive and high attention commentary about the Vermont Senator. Beyond this, despite it coming up as a topic, post-debate narratives don’t seem to be organized around the wealth/income inequality issue in the same way that media before has been.
  • Here are the five most on-narrative takes published about the first debate:

ET Election Index: The First Debate

This content is related to the Epsilon Theory Election Index, a series we introduced here in hopes of better informing citizens and voters about the political narratives present in US national media.


With one debate down (and another to come tonight), we take a brief step back to look at how media are telling the story about the debate and its participants. As usual, we start with the network graph from our queries of national and local media, blogs and magazines published today about yesterday’s debates.

Source: Quid, Epsilon Theory

Our takeaways from this coverage:

  • In general, coverage has been pretty negative. Format, technical difficulties, reviews of individual performances, single-issue opinion writers upset that their issue didn’t get enough airtime, etc. Not unusual for debates to have these logistical criticisms pop up, but there really weren’t any positive language clusters. No shining stars or issues that really popped out consistently in media coverage.
  • The one possible exception might be Tulsi Gabbard, whose dedicated coverage has been more on-narrative than her pre-debate coverage has been at any point. While not polling in a place to make her a credible challenger at this point, we think (our opinion, not fact) that her willingness to speak brashly on military and police action is likely to be a continuing point of provocation for moderate, traditional Democratic candidates.
  • Other than round-up pieces, the most central clusters to the narrative emerging from the debates include language about (1) border detention, (2) gender and equality and (3) an economy that “isn’t working for everyone.”
  • As we’ve observed elsewhere, the narrative that continues to resonate and is most often and loudly repeated by missionaries in media uses the language of economic socialism and intense emphasis on the situations of poor and minority citizens.
  • Other wonkish topics and policies may matter to voters, but they aren’t – and we think, won’t – be the issues that come before citizens through the media.
  • Warren-relative language was ubiquitous throughout the clusters – so much so that there isn’t really a Warren cluster. They are ALL Warren clusters.
  • That doesn’t mean the coverage was universally positive or cohesive. Our general sense is that most media narratives were mixed on her performance, at best. Most used this as an opportunity to present some of the lower-polling candidates as “surprises” or “underdogs.”
  • As cringe-worthy and meme-worthy as many (including comedians and late-night hosts) found Beto’s unprompted Spanish response to be, coverage of it is peripheral to the narratives emerging from the debate. We have other reasons for concern about the trajectory of his candidacy, but we are sellers of the idea that this is a deadly issue.
  • Same goes for technical difficulties – both are just being treated as quirky side issues.
  • Here are the five most on-narrative takes published about the first debate:

Election Rewind: June 2015

One of the questions that has come up from our ET Live! discussions about the 2020 elections is, “What would this have showed at this point in the last election?” It’s a good question, so we decided to take a look.

It so happens that May 2015 was a pivotal month in the GOP primaries. Now-president Trump was polling at low single digits, but that wouldn’t last for long. So what would you have seen in narrative structure that month? First, when looking at Attention – our measure for the external similarity of candidate-specific language used in media with the language relating to the election more broadly, you would have seen a few candidates who, despite widespread popularity, suffered from a narrative structure at odds with what the 2016 election was perceived to ‘be about.’ You also would have seen two candidates with middling single-digit poll numbers whose narratives in were media very in-line with what it was perceived to ‘be about.’

What about sentiment?

There is a tendency when we think about elections and politics to think about populations and what they care about. We build up predictive engines based on demographics, expressed preferences and revealed preferences on issues, responses to poll questions and other inputs. In general, those models are pretty good. Well, they’re OK.

We think that the time has come to stop predicting and to start observing – not what polls are saying or what our network of friends or social media followers are retweeting, but what the crowd thinks that the crowd thinks. We continue to think that the widest possible net of mass media, blogs and other primary content gives us the best window into observing that thing, which we call Common Knowledge.

Does this mean that Biden is Jeb, and that we can hand out the Carson, Cruz and Trump roles to some combination of Warren, Harris and Sanders? No. There are different circumstances for each (not least Sanders’s prior electoral history), and there are a lot of actual events that can still influence that common knowledge. But for now, what we can observe is that Biden is a poor fit for what everyone knows that everyone knows the 2020 election is about. We can similarly observe that Sanders in particular is a good fit for that common knowledge, with a hell of a lot of rah-rah cheering coming from the kinds of language used to cover him by the impartial media.

It means something for the way we consume information, if we want our own views to be less shaped by our intuitive ability to unwittingly internalize and co-opt that common knowledge as our own points of view.

It also means something for how we ought to anticipate primary season playing out.

Warren in June: Back in an Unflattering Spotlight


Warren Narrative Map as of May 31, 2019

Source: Quid, Epsilon Theory

Warren Narrative Commentary

  • The lower attention level of Sen. Warren’s narrative compared to that of other top candidates is, we think, indicative of two patterns.
    • The first is that Warren’s campaign and its coverage have emphasized her role with respect to issues that are less relevant to overall election narratives. In particular, impeachment of President Trump continues to be an issue linked in media to Warren, but it is treated as a peripheral election issue in media, at best.
    • The second is that, even with respect to otherwise high attention topics like student loans, socialist or highly progressive policies, etc., the language used in coverage of Warren’s policy ideas treats her association with them as almost ‘accidental’.

  • In a sense, our opinion is that Sen. Warren’s attempts to balance support of centrists while still claiming the mantle of the ‘very progressive’ candidate – despite a track record that generally supports the latter contention – are not yet resonating in political media.

  • They are resonating more than Biden’s narrative, and we would not be surprised to see Warren take some of his early support.

  • As Sen. Warren’s very prominent senatorial activities take a backseat to campaigning over the coming months, we will be interested to see if the overall cohesion of and attention to her platform and candidacy improves from today’s somewhat fractured narrative.

  • If it is true that Warren’s best opportunities lie ahead in the more wonkish phases of the primary season, we would also counsel monitoring the sentiment of her coverage, which, while not as abysmal as Biden’s, has been increasingly negative over the course of 2019.

  • The leading women candidates – Harris and Warren – have both been treated to far more negative language in political media in recent months than all of their non-Biden counterparts. On that basis, we counsel care in consuming potentially affected / opinion-influenced news content about these candidates.

  • Topically, Warren appears to have attempted to carve herself out as THE anti-trust, anti-corporate power candidate, and as the solving-student-loans candidate. She’s been prominent with her plan on the opioid epidemic, too. But the media have not yet fully embraced the relationship between these topics and Warren’s policies. In other words, when they write about ‘the student loan crisis’, they only occasionally refer to Warren with language about her proposals. However internally coherent the platform may be, the link isn’t nearly as connected in media election narratives as, say, Sanders and healthcare.

  • These are both important issues to 2020 election narratives. We think closing the gap on her association with these issues would have a significant impact on the common knowledge promoted by media about her candidacy.

  • The Native American claims issue is still there, albeit in mostly right-leaning media. It also remains surprisingly central to discussions of Warren. But it has declined substantially in quantity, attention and cohesion over the last few months.

Warren Narrative Attention as of May 31, 2019

  • Whatever Warren Narrative exists is roughly average among candidates in terms of its consistency with broader election narratives and language.

  • This does, however, represent a meaningful increase over the last few months as campaign dialogue has moved away from new candidate announcements and back toward more issue-focused language that has tended to align well with Warren’s own coverage.

  • In fact, Warren trailed only Kamala Harris in our attention measure for the month of May alone. Unfortunately, that higher attention has not been accompanied by positive coverage.

Warren Narrative Cohesion as of May 31, 2019

  • Senator Warren’s narrative has been very consistent (note the tightness of the Y-Axis scale on the above graphic) throughout early 2019, and ranking in the top half of candidates but soundly behind that of Sen. Sanders.

  • The internal consistency of language used to describe Sen. Warren and her campaign has been moderate, but is probably being suppressed somewhat by her very public involvement in the US Senate and the added complexity of non-election issues on which she makes public statements.

  • Regardless of the causes, as with most other candidates, a Warren Narrative has not yet crystallized in media in the same way that a Bernie Sanders Narrative (or for that matter, a Joe Biden narrative) has.

Warren Narrative Sentiment as of May 31, 2019

  • While the data we are working with can’t ascribe intent, we do note that each of Harris, Warren and Gabbard have been soundly in the bottom 4 of sentiment for the second quarter of 2019 (joined only by Biden).

  • As important and central as sex and gender language have been thus far, it is striking that media accounts of the three most prominent woman candidates have tended toward such negative language in coverage.

  • We would be mindful of this in reading all of the coverage of these candidates, but in the case of Warren in particular, whose descent in sentiment has been so consistent, we would be doubly cautious of the influence of the revealed preferences, if you will, of authors and news outlets.

  • We also note that this has happened even as the broadly negative coverage of her prior comments about Native American ancestry have faded. This sentiment swing does not to be a feature of right-leaning publications alone.

Sanders in June: Polarizing…Except in Media


Sanders Narrative Map as of May 31, 2019

Source: Quid, Epsilon Theory

Sanders Narrative Commentary

  • Of all major candidates, Sanders coverage demonstrates the highest attention, the highest cohesion and the highest sentiment (although sentiment relating to coverage of Pete Buttigieg, who has emerged somewhat in May and June, rivals that of Sanders).
  • The high cohesion – visualized by the overall compression of distance between nodes and clusters in the graph above – indicates high similarity of language across articles referencing Sen. Sanders. Authors’ references to Sanders follow a structured taxonomy and almost obligatory references to his age, socialism-related terms and his 2016 campaign.
  • We think this alone justifies significant caution when reading news about Sanders and his campaign. We think the gravity of the strong Sanders narrative has the capacity to skew the language and implicit/explicit conclusions included in news reports in the direction of that narrative.
  • Discussions of democratic socialism, the green new deal and related topics concerning universal health care, workers rights and unions, taxing the rich, etc. all occupy the most central, most connected part of Sanders’s narrative. At this point in the campaign, the Sanders narrative is not only more cohesive than that of other candidates, it is cohesive around key issues.
  • Those key issues and the language used to discuss them happen to be the ones that the media have been most engaged in writing about – popular democratic socialist policies are central to the narratives of the Democratic primary – which is the driver of Sanders’s very high attention score in our system.
  • There are only two central clusters within the Sanders graph that are remotely negative: the first includes articles defined by ‘old white guy’ language, usually expressed with concern about dual standards in discussions of electability and likability of Sen. Warren and Sen. Harris, in particular. The second are discussions of Sanders’s remarks on Israel and Netanyahu.
  • From a ‘predictiveness’ perspective, we think it is more difficult to judge how much the influence of media will be on a candidate like Sanders, who has a history and has established “opinions” from many likely voters. Still, the media is (intentionally or unintentionally) promoting common knowledge that “Sanders is the candidate with the policies that will make America America again.” We are less convinced than some that voters still turned off by the 2016 primary or by Sanders’s socialist credentials won’t be affected.

Sanders Narrative Attention as of May 31, 2019

  • Attention of coverage on Sanders declined somewhat in March with the entrance of additional candidates; since that time, however, it has remained stable and still well above that of nearly all other candidates.
  • For most of 2019, populist and democratic socialist-related topics and language have dominated election coverage, which we suspect is the main reason for Sen. Sanders’s attention measurement under our methodology.

Sanders Narrative Cohesion as of May 31, 2019

  • With the exception of the period prior to his formal announcement of a 2020 campaign, the cohesion of coverage of Sen. Sanders has been both very high and consistent.
  • We think that the narrative of Sanders – the ‘socialist’ candidate with a very progressive policy history – is crystallized across media coverage, causing all pieces to veer toward mentions of those issues or the use of language consistent with that common knowledge, even on disparate topics.
  • As noted previously, we believe the presence of such a cohesive narrative warrants extra caution when reading news about Sanders (from all parties, regardless of perspective on his candidacy).

Sanders Narrative Sentiment as of May 31, 2019

  • Interestingly, the sentiment attached to Sanders coverage has steadily increased in every month of the primary campaign.
  • This is an unusual pattern, and isn’t reflective of anything happening more broadly with the mix of topics over these five months – it certainly isn’t a pattern shared by any other candidate.
  • It is, however, only five months. This rise could absolutely be a feature of some random relationship of Sanders to topics that were covered more positively. Still, as consumers of news, we would be concerned that this might be indicative of knowing or unknowing preferences on the part of journalists authoring stories or editors assigning coverage.
  • Almost universally, we think consumers of political news – supporters and opponents of Sen. Sanders alike – should be asking “Why am I reading this now?” for almost every Sanders article they read.

Biden in June: Popular but Disconnected

This candidate-specific report is part of our Epsilon Theory Election Index series, in which we turn the tools of narrative analysis and natural language processing to media coverage of the 2020 election. For more about this series, including explanations of all of the terminology and measurement used here, please read our first installment here.


Biden Narrative Map as of May 31, 2019

Source: Quid, Epsilon Theory

Biden Narrative Commentary

  • In general, while sentiment attached to Biden by the political media continues to be far more negative than for other major candidates, some of that sentiment is isolated to topics that don’t appear to be getting much traction: his role in the 1994 Crime Bill, the Trump/NK ‘low IQ’ banter and ‘Ukraine scandal’ being chief among them.

  • Articles about these topics tend to be standalone and linguistically distinct. They are usually only about that thing, and aren’t routinely being referred to in ‘core’ articles about the election and Biden’s candidacy. We judge that common knowledge at this point is that ‘these don’t matter and aren’t relevant to the election.’

  • That isn’t true, however, for Biden’s history with Anita Hill. It is much more connected to Biden’s central narrative, and the topics of sexism and racism – especially in context of the Democratic Party’s role in expunging them – are central to broader election narratives in a way that make this the ‘negative’ issue of concern.

  • The most striking feature of Biden coverage, however, is that the most connected, most on-narrative articles aren’t about Biden at all. They’re about Donald Trump or about a policy proposal or event from another candidate. In other words, Biden’s narrative is largely being defined as a measuring stick or comparison (e.g. will this allow candidate X to close the gap with Joe Biden in polls?) without much in the way of primary influence.

  • There is one exception: there DOES appear to be common knowledge that Biden is the candidate who can secure more votes from the black community.

  • Other than that, as expressed elsewhere our view is that Biden at this point is a candidate whose narrative in media is moderately well-defined, but mostly in terms of things other than his policies or candidacy, and whose narrative is at odds with the issues being promoted as ‘important’ to the 2019 election.

  • We suspect that the realization of this treatment – whether justified or not – has prompted some of Biden’s more aggressive recent moves, including his flip on the Hyde Amendment and declaration of Donald Trump as an ‘existential threat.’

Biden Narrative Attention as of May 31, 2019

  • As noted previously, relative to the other major candidates, the Biden narrative is very much at odds with the broader 2020 election narrative, which is what we track through our Attention measure.

  • What people write, care, and think about when they write, care, and think about the 2020 election is very different from what they write, care, and think about Joe Biden and his candidacy.

  • This effect has been far more pronounced since his announcement.

Biden Narrative Cohesion as of May 31, 2019

  • While attachment to broader election narratives has been weak, relative to other major candidates, the internal consistency of Biden coverage is comparable and competitive.

  • As his campaign has ramped up – and as his narrative as “front-runner” has emerged – the cohesion of a Biden narrative has accelerated.

  • In short, we see a moderately cohesive network of articles that simply isn’t all that related to anything else being written about the election, key topics, or policy issues.

Biden Narrative Sentiment as of May 31, 2019

  • While the sentiment attached to language used in Biden coverage has bounced back slightly from the lows following his candidacy announcement, it remains far below that of any other candidate over and during almost any period.

  • Some of this can be attributed to the early oppo research effects faced by any front-runner with an established legislative and political history, but even so it should be noted that this coverage has resulted in the presumably desired negative sentiment effects.

  • One could argue that Biden has actually been somewhat fortunate. The scandal and criticism-related topics other than Anita Hill have been sufficiently distant and disconnected from most coverage that their discussions have not really impacted the broader sentiment. As you can see from the sentiment visualization of the Biden narrative map for May 2019 below, the two most negative sentiment clusters are among the least well-connected to the broader Biden narrative.
  • Still, it remains our view that broader election media has (intentionally or unintentionally) treated socialism, green new deal, equality, difference from Trump and other issues outside of Biden’s wheelhouse as The Issues of the 2020 election thus far, and has been unfavorable in treatment of candidates, centrist or otherwise, who have not incorporated those principles into their own story-telling and platforms.

  • Biden’s more recent attempts to appeal to progressives – and the media response to them – will be telling and should be monitored, but our conclusions continue to be that (1) we would read any news about Biden with significant caution about the underlying attempts of the outlet to establish common knowledge about his candidacy, and that (2) we would not be surprised to see continued erosion in support for Biden in favor of Warren, Sanders and (to a lesser extent) Harris.

The ET Election Index – May 2019

This is the second installment of Epsilon Theory’s new monthly feature – the ET 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 goal isn’t to uncover ‘media bias’.

Our goal isn’t to discuss the ‘fairness’ of coverage to different candidates.

Our goal isn’t to ‘fact check.’

Our goal is to make you a better, more informed consumer of political news by showing you indicators that the news you are reading may be affected by (1) adherence to narratives and other abstractions, (2) the association/conflation of topics and (3) the presence of opinions. Our goal is to help you – as much as it is possible to do – to cut through the intentional or unintentional ways in which media outlets guide you how to think about various issues, an activity we call Fiat News.

Our goal is to help you make up your own damn mind.

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.

Election Narrative Structure as of May 31, 2019

Source: Quid, Epsilon Theory

Commentary on Election Narrative Structure

  • We observed two major focal points within the 2020 Election narrative structure in May: an ongoing peripheral focus on impeachment discussions and election security / infrastructure, and a much more centralized discussion of the current roster of candidates and prominent election issues for 2020.
  • The five most influential clusters on the overall network – meaning those with the most universal language similarity – were (1) Trump Super PAC, Embraces Big Money, (2) Electability, Likability, Sexism, Biases and Voter Suppression, (3) Markets Respond to Escalating China Trade War, (4) Joe Biden Beats President in Polls and (5) Strong Trump Economy
  • Not far behind these topics and very close to the center of the overall network is the large cluster defined by references to Climate Change, the Green New Deal and terms associated with Socialism. The combination of these terms with discussion of sexism, biases and voter suppression leads us to believe that the core election narrative promulgated through media continues to be well-left-of-center populism.
  • With the candidacy announcements of most major candidates now firmly in the past, most topical clustering has reverted to election issues and platforms rather than individual candidates.
  • Our separate research on China Trade War narratives indicates that the “existential”, “national security” language has been increasingly conflated with trade war issues, especially beginning in May 2019. Tackling this issue and using it will be a unique challenge for Democratic candidates, some of whose support base is ambivalent (at best) about free trade.
  • Whether it will match up with the final campaign remains to be seen, but we believe current media narratives are (by and large) positioning President Trump on the basis of his economic agenda and Democratic campaigners on their association with social / equality / fairness issues.

Candidate Cohesion Summary

Commentary on Candidate Cohesion

  • Candidates in bold are those for whose coverage we would exercise special caution in our news consumption habits.
  • Coverage of Sanders remains by far the most cohesive. Outlets have established language patterns and terms for discussing Sanders and his policies, and whether the authors (whether opinion or news writers) have an opinion, the abstraction of Sanders into ‘the Socialist Candidate” appears to be ubiquitous. We would read any Sanders coverage with special attention to this tendency.
  • Cohesiveness around a Biden narrative has broken down somewhat after the fever pitch following his candidacy announcements. We would always counsel caution when consuming political news, but no longer think there is cause for particular caution with respect to Biden news.
  • The deterioration of any coherent narrative about O’Rourke and Booker continued in May. They still both receive significant volume of coverage, but the language used across outlets was highly variable and disparate in tone and approach, and lacked any abstractions of the candidates or policies into any ‘idea’ or ‘role’ in the primary process. We think it is fair to say there is no longer a Beto or Booker narrative.
  • The biggest cohesion riser – and we think, the major beneficiary to Beto’s decline – is Pete Buttigieg. Earlier media reports were much more disparate, but Mayor Pete’s May coverage nearly approached Sanders-level unanimity among media outlets.

Candidate Sentiment Summary

Commentary on Candidate Sentiment

  • The candidates in bold are those whose levels or changes in coverage sentiment would give us pause in our consumption of election coverage.
  • Yang’s extremely positive coverage continued in May, but on only a slight increase in volume.
  • While coverage of Biden was less negative in May – mostly on the back of a strong cluster of stories written about polls demonstrating his advantage against Donald Trump – it remained below average and was not enough to pull YTD sentiment out of last place among major candidates.
  • In addition to a clear cohesive narrative, Bernie Sanders enjoyed the most positive coverage of all major (non-Yang) candidates in May.
  • Gabbard, Harris and Warren continued to be associated with far more negative language, in spite of a high concentration of articles highlighting and empathizing with the existence of a potential double standard in the evaluation of female candidates.

Candidate Attention Summary

Commentary on Candidate Attention

  • As was the case in our April update, among high-polling candidates, the broader election narrative (i.e. topics, language, people and issues) aligns most closely with Sanders and has since the beginning of the primary process. In other words, the things that journalists, opinion writers, bloggers and pundits are saying about the election and its major issues are the things they also associate with Bernie Sanders.
  • The language used in discussions of Booker and Klobuchar continues to be the mirror image of Sanders: completely disconnected from broader election narratives.
  • While a more coherent Buttigieg narrative emerged in May, it was similarly disconnected from broader election narratives. Like Gabbard or Yang, Buttigieg runs some risk of having clear stories told about him which have little to do with the central stories being told about the election.
  • Despite the huge volume of his coverage, Biden’s attention dipped below that of Warren, Harris and Sanders in May.
  • In general, our interpretation is that media treatment at this stage has been more aligned with more progressive candidates and their platforms and less favorable to more centrist candidates and theirs.
  • The goal of our research is to equip voters to consume election information with clear eyes, and not necessarily as a predictive science. Given this narrative structure, however, we would not be surprised to see continued erosion and splitting of Biden’s current support among Warren, Harris and Sanders.

The Continuing Series

Following the upcoming candidate-specific reports, this series – including issue-related narrative analyses – will continue as an Epsilon Theory Premium feature and will require a subscription.

If you are looking for a more detailed package of our election narrative signals and analytics, including raw data and candidate- and issue-level narrative structure analysis, please email us directly at info@epsilontheory.com.

The ET Election Index – April 2019

In celebration / abject dread of the 2020 election cycle that is already upon us, Epsilon Theory is beginning a new monthly feature – the ET 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 goal isn’t to uncover ‘media bias’.

Our goal isn’t to discuss the ‘fairness’ of coverage to different candidates.

Our goal isn’t to ‘fact check.’

Our goal is to make you a better, more informed consumer of political news by showing you indicators that the news you are reading may be affected by (1) adherence to narratives and other abstractions, (2) the association/conflation of topics and (3) the presence of opinions. Our goal is to help you – as much as it is possible to do – to cut through the intentional or unintentional ways in which media outlets guide you how to think about various issues, an activity we call Fiat News.

Our goal is to help you make up your own damn mind.


What do we do?

We leverage the natural language processing (NLP) tools from Quid to construct network graphs of all available sources from the LexisNexis Newsdesk database. These network graphs are built by comparing each word of the language used in each queried story against each word in every other story, resulting in a matrix of similarity between each such article.

Here at Epsilon Theory, we then take this output and attached meta-data to build and score a range of signals by topic – which can be either candidates, policies or any other categorization scheme which fits sensibly inside election coverage.

Any prose you see under the heading “commentary” you should read as our judgments and opinion. Everything else – unless otherwise noted – you should take as the systematic output of the process and signals described above.

Cohesion: Our Measure of Narrative-Adherence

We think that news and opinion journalism often replace facts and statements with abstractions of those facts. Those abstractions may take the form of euphemisms, logical frameworks, conflations or descriptive turns of phrase. Over time, especially when the abstractions are emotionally or culturally powerful (memes), these abstractions can become increasingly divorced from the facts they were originally built to represent. They may also become loaded with other, more subjective, meaning. This makes interpretation both difficult and variable by individual.

We call these abstractions narrative. By measuring the aggregate similarity of language used by writers in the discussion of a topic over time, we believe we can track the extent to which a narrative exists for that topic. We call this Cohesion.

We also believe that a cohesive narrative tends to cause future content to adhere to the conventions and language that have become common knowledge for that topic. As a result, when we consume news content about topics or individuals with high Cohesion, we must be aware that the information may be presented to us in a way that aligns it with existing narrative abstractions rather than on an independent basis.

In effect, this is one of the ways in which we are subtly guided how to think about an issue.

What does Cohesion look like?

Our measure is based on the underlying aggregate similarity between all of the articles relating to a single topic. When more articles refer to Pete Buttigieg as “a gay, millennial” candidate, this measure increases. When more articles refer to Elizabeth Warren as “policy-oriented”, this measure increases. When articles vary between isolated references to Andrew Yang’s UBI proposal, or his penchant for cryptocurrency, or descriptions of him as an “entrepreneur”, or “technologically savvy”, that cohesion measure will typically fall.

While we calculate our measure based on the underlying data describing this aggregate similarity, we also present visualizations from our partners at Quid. We think it can be helpful to see what connections – and the lack of connections – look like. An extremely low cohesion network might look something like the below.

Source: Quid, Epsilon Theory

In contrast, a high cohesion network might look like the below. In both of these graphs, similarity is presented by proximity on the chart, emphasized in very high cases (‘adjacencies’) by a connecting line, and organized by color into clusters of topical or linguistic meaning/similarity.

Source: Quid, Epsilon Theory

Attention: Our Measure of Association and Conflation of Topics

We also believe it is useful to note when authors use increasingly similar language to describe two topics, especially when one topic is a sub-set of the other. We think that this behavior is often an expressed desire (or belief) on the part of the author to establish or presume a relationship between those topics. Because the relationship between topics and/or people forms a significant part of how we think about issues, their conflation is an effective way to guide us in how to think about a topic.

In other words, these are the topics which warrant the question: “Why am I reading this NOW?”

We call our measure of the similarity of language used between two topics Attention.

What does Attention look like?

While visualization of Attention doesn’t fully capture how connected a topic is to (or better, within) another, it can still be useful to develop an intuition for it. A high Attention relationship might look like the below, where highlighted nodes and connecting lines identify those of a topic within another (in this case, references to the Dallas Cowboys within broader NFL coverage):

Source: Quid, Epsilon Theory

In contrast, a low Attention topic would be one for which the centrality and connectivity of articles was limited. The below visualization presents what that might look like (in this case, an illustration of references to concussions within the same broader NFL coverage):

Source: Quid, Epsilon Theory

Sentiment: Our Measure of the Potential Influence of Author Opinion

We also believe that the graying of lines between news and non-news coverage has created an environment in which nearly every piece can be expected to include at least some affect of the author or publisher, even when great pains are taken by fastidious, ethical journalists to avoid it.

We think sentiment measures – even ours – should be viewed with caution. The main reason is that many topics are intrinsically related to terms which have ‘negative’ sentiment under any scoring system. You won’t find many articles about climate change with a cheery tone, but that doesn’t mean that the author is necessarily including opinion. What can be useful, however, are comparisons of sentiment over time, or comparisons of like coverage – for example, across candidates. That is where we tend to focus our efforts.

Who is it for?

People who want to make up their own damn minds.

People who want to know when conscious or subconscious use of language by politicians, media members and other influential figures and entities is influencing the lens through which we think about and – with other civically and politically minded fellow citizens – collectively experience the debate about our country’s political direction.

Some outlets try to approach these questions through analysis with sentiment, or with sentiment and time-on-air measures, or with fact checks. Those are all adequate for some purposes, but they don’t capture the things we intuitively sense about the nature of political coverage and the stories we begin telling ourselves about “what this candidate means for America” or “what it means that this policy is even being discussed.” Those kinds of abstractions are where we tend to be led toward politically polarized outcomes. They’re also the kind of thing that NLP is very well-suited to explore.

With that, here’s our first look at the summary of Election Narratives as of April 30, 2019, which we will supplement with Candidate and Policy-specific detail in future pieces.


Election Narrative Structure as of April 30, 2019

Source: Quid, Epsilon Theory

Commentary on Election Narrative Structure

  • The Democratic rallying cry that has resonated throughout election coverage is, roughly, “This is not America.” Right now that looks like a progressive populist message, caught up in a variety of discussions about and coverage of socialism, freer immigration, rural America, student debt, and Latino turnout.
  • Among the large clusters of articles that are not “round-up” style pieces, the five most influential clusters on the overall network – meaning those with the most universal language similarity – are (1) Latino Turnout, (2) Mayor Pete Buttigieg, (3) Mueller Report, (4) The Joe Biden Allegations and (5) Rural America.
  • Those with the most similarity to multiple different topics – indicating strong external relevance to other subjects – are (1) New Hampshire Democrats, (2) Latino Turnout, (3) Student Loans and Marijuana, (4) Not America, Ilhan Omar and Socialism, and (5) Mueller Report.
  • While Biden’s coverage was much higher in volume than that of other candidates in April, the content that was most similar to the overall story being told about elections was not about Biden. His most connected topics in April were, alas, not related to his candidacy but to behavioral allegations.
  • PTSD from 2016 seems to have kept Rural America squarely in the crosshairs of a lot of the media discussion about the 2020 election.
  • While an electoral strategy predicated on Latino turnout has not produced much content (yet), what content is there is very consistent with the overall story being told by the media about the 2020 election.

Candidate Cohesion Summary

Commentary on Candidate Cohesion

  • Candidates in bold are those for whose coverage we would exercise special caution in our news consumption habits.
  • Even pre-candidacy, coverage of Biden used very consistent language. With his formal announcement, the April measure placed him at the top. While some of this is related to the topical consistency of the announcement itself (i.e. we don’t think everyone using and printing the same quotes from the announcement is necessarily concerning), we still think that readers should be mindful of adherence to media narratives in coverage of Biden’s candidacy at this time.
  • We share a similar view of Sanders, whose prior candidacy and fairly well-established ‘story’ has created the strongest primary-season cohesion of all candidates. The media have largely decided how to report on Sanders and how to structure their coverage of events relating to him. We continue to recommend caution to readers.
  • Yang, Gabbard, Buttigieg and Klobuchar have tended to yield the least internally consistent language among all candidates; however, both Buttigieg and Gabbard saw meaningful gains in April. We think that is indicative of Buttigieg’s formal candidacy announcement and crystallization of “anti-War” coverage for Gabbard, but readers should be on guard.
  • We are observing some deteriorating narratives for Warren and O’Rourke. That doesn’t mean a negative – it simply means that media articles are not as uniformly on-message in the language they use to describe each of these candidates. We think this is more ‘real’ in the case of O’Rourke, where the previously generous coverage has begun to break down. In our judgment, the weakened narrative structure for Warren is more the result of regular involvement in non-election news creeping in and diversifying the focus and language used.

Candidate Sentiment Summary

Commentary on Candidate Sentiment

  • The candidates in bold are those whose levels or changes in coverage sentiment would give us pause in our consumption of election coverage.
  • Yang’s sporadic coverage is almost universally glowing, as he has thus far received disproportionately positive “Yang Gang”-style puff piece features. As always, we aren’t arguing intent to make him a more prominent candidate, but we would expect his coverage, cohesion and popularity to grow on this basis.
  • Biden gets the other side of the coin. Coverage of him is universally – in both April and on a primary-season-to-date basis – decidedly more negative than that of other candidate. It is somewhat harder to ignore the media’s displeasure with his polled popularity as a candidate.
  • We would use caution in our takeaways from the big movers as well. Before his announcement, Buttigieg was a Yang-like darling in the media. Afterward, he was subjected to substantially more critical language.
  • Sanders, on the other hand, has been a net beneficiary of negativity toward Biden in particular, at least in media sentiment, if not in polls. As noted in the cohesion section, we would exercise caution in consuming news and feature content about both candidates at this time.

Candidate Attention Summary

Commentary on Candidate Attention

  • Among high-polling candidates, the broader election narrative (i.e. topics, language, people and issues) aligns most closely with Sanders and has since the beginning of the primary process. In other words, the things that journalists, opinion writers, bloggers and pundits are saying about the election and its major issues are the things they also associate with Bernie Sanders.
  • Both Biden and Buttigieg’s alignment with key election narratives fell sharply in April, although this is likely because most coverage in April focused on their candidacy announcements (which tend to be unrelated to broader election issues). As noted above, one Buttigieg-related cluster (not the full range of articles relating to him) sat nearest the ‘Zeitgeist’ of the election coverage in April, so ‘non-announcement’ coverage still seems to be high attention for Mayor Pete.
  • By analog, the narrative alignment of Beto’s candidacy improved after a similar drop he experienced in March, although in the aggregate his issues-and-language connection to the broader election coverage is still lower than it was at the beginning of 2019.
  • Despite his popularity in polls, the language used in Biden articles is broadly at odds with general election narratives (which, as you might imagine from Sanders’ much higher Attention score, is because comfort discussing socialist policy is very much on-narrative). It will be worth noting in May how much the announcement drove Biden’s figure lower.
  • In general, our interpretation is that media treatment at this stage has been more aligned with more progressive candidates and their platforms and less favorable to more centrist candidates and theirs.

The Continuing Series

We will publish candidate-specific reports from April for the rest of May as part of our free service on the Epsilon Theory website. Thereafter, this series will continue as an Epsilon Theory Premium feature and will require a subscription.

If you are looking for a more detailed package of our election narrative signals and analytics, including raw data and candidate- and issue-level narrative structure analysis, please email us directly at info@epsilontheory.com.