Media bias in the Democratic primary
There are two types of bias at play in the media, and both contribute to unfairness: Bias in tone, and bias in volume. For example, an…
There are two types of bias at play in the media, and both contribute to unfairness: Bias in tone, and bias in volume. For example, an op-ed attacking Beto O’Rourke’s use of a Latino nickname is biased in tone; an infographic of candidate fundraising including all the top fundraisers except for Andrew Yang is an example of bias in volume.
Both types of bias were very visible in 2016, but it’s much easier to establish bias in volume than bias in tone, so I will be concentrating on bias in volume, mainly, using the data available from FiveThirtyEight. Unless otherwise noted, average or aggregate poll numbers are generated as in this article. There are significant differences in the tone of coverage of different candidates even within “blue” media outlets generally friendly to Democrats, and my current impression is that blue media outlets favor less electable candidates, particularly Elizabeth Warren.
I will present several measures of media bias, most based on comparing polling numbers to media coverage.
Which candidates have gotten more media coverage and why?
Media bias in volume is both widespread and unfair. However, the equal time rule was gutted a long time ago, and candidates receive wildly different amounts of coverage. The DNC has tried to give a few public nods to the idea of treating candidates who are on the debate stage as being on equal footing with one another, but candidates are not receiving equal time even on the debate stage.
Comparing the two Joes is very striking. Joe Biden gets more coverage in a week than Joe Sestak has in the past year. However, when we consider the difference in their polling numbers, fundraising, and endorsements, the media being biased towards coverage of Biden as the frontrunner is understandable.
On a certain level, it isn’t fair; but the unfairness can be grounded objectively in the fact that Biden is far better known and far more popular than Sestak. This is a horse race bias in favor of front-runners and against struggling contenders, rather than a bias against Sestak or in favor of Biden.
Comparing media coverage and polling
The answer to most of the differences in coverage is polling. I contend that what’s left after that is media bias for and against specific candidates.
Media coverage both drives and is driven by polling, so the two correlate strongly in the long run. Some of the anomalies can be explained by breaking down the data from month to month, such as Tom Steyer’s position on the chart as a late-entering candidate. However, other anomalies remain unexplained.
From this big picture perspective, it seems clear that the media has favored Elizabeth Warren over Bernie Sanders, Cory Booker over Beto O’Rourke, and many other candidates over Andrew Yang.
Within the backfield of lower-polling candidates, moderates from inland states (Michael Bennet, Steve Bullock, John Hickenlooper, and Tim Ryan) have been overlooked in favor of politicians from coastal states (Kirsten Gillibrand, Jay Inslee, Eric Swalwell, and Bill de Blasio). However — and I will repeat this point — down in the < 1% range, it’s hard to have any confidence in any meaningful differences in polling averages.
If we look more carefully at the numbers over time, they usually correlate both within any given week and any given month. “Horse race” coverage is widespread, and most candidates get similar levels of coverage per point of polling most of the time.
Some of the individual exceptions in a particular week or month can be explained by news events. To avoid concentrating on the effect of any singular news event, I’ll be looking at the relationship between coverage and polling in a median, or typical, week. When the same candidates are repeatedly and consistently over- or under- covered relative to their polling, week after week, media bias is the simplest explanation.
The relationship between polling and coverage is not exactly proportionate; it tends to deviate at the higher and lower ends. The natural answer to this question is to do a linear regression.
This shows us two slightly different but related measures of of media bias. The linear model is preferable as it takes into account more information and doesn’t involve dividing by numbers that approach zero.
Looking at both measures, it’s clear that Beto O’Rourke has been shortchanged, with the second-lowest ratio of all candidates in spite of averaging only seventh in the polls. We can similarly highlight Andrew Yang, Steve Bullock, Joe Sestak, and Mike Gravel as routinely under-covered relative to their polling.
On the side of positive coverage bias, the media has preferred to cover Elizabeth Warren, Cory Booker, Kirsten Gillibrand, and Bill de Blasio. Perhaps remarkably, Pete Buttigieg seems to be covered exactly the appropriate amount using either measure. For all other candidates, the indicators are more mixed; we can make a few relative determinations between the other candidates, but it’s not clear they experience either net positive or net negative bias in their volume of media coverage.
Candidates as competitors for coverage: Evaluating every case
There are a couple of major issues we’ve touched on here. One is that polling and coverage have a mostly — but not entirely — proportionate relationship. The anomalies we can identify confidently are based on pairwise comparisons between candidates who have relatively similar polling numbers. Fortunately, we can take all of those anomalous pairwise comparisons and use them to generate a rating of the direction of media bias relative to polling.
As of my analysis, there are 921 cases in the data where a candidate received less coverage than another candidate over the course of a week, in spite of the fact that they were polling better both during that week and the week before. There’s a great mathematical tool for teasing out the relative strength of competitors from a sequence of pairwise competitions: It’s called an Elo rating, after physicist and chess player Arpad Elo.
Running the 921 pairwise comparisons through the Elo rating formula gives us a measure of unexplained media preference for (or against) a candidate.
There are two candidates whose ratings indicate they are very strongly snubbed by the media: Andrew Yang (rated 675) and Beto O’Rourke (rated 836).
It’s not clear if third place properly belongs to Bernie Sanders (1022) or Joe Biden (958). Since Joe Biden has been the polling front-runner every single week of the race, his rating is artificially depressed by this measure; he can only “lose” to candidates with lower polling numbers. Moderate politicians as a group tend to be below average. Surprisingly, this includes Pete Buttigieg (1068), who looks close to average in other views of the subject.
The notable candidates with high ratings include Kirsten Gillibrand (1408) and Elizabeth Warren (1335); of the candidates still in the race, Elizabeth Warren is a clear media favorite. The top rated candidates by this method are Eric Swalwell, Bill de Blasio, Jay Inslee, and Seth Moulton. It’s worth asking — again — what the scale of importance is?
How many stories and clips are these candidates gaining and losing?
If we take the 921 pairwise comparisons in the data set and pare them down to each week, we can identify a minimum shift in coverage that would eliminate any apparent snubs. Doing this puts the positive volume bias into perspective: Seth Moulton may have been covered instead of other obscure candidates, but very little coverage was at stake.
Cory Booker stands out by this measure, surpassing the usual suspects of Kirsten Gillibrand and Elizabeth Warren. The most snubbed candidates have again clearly been Beto O’Rourke and Andrew Yang.
A small note on the geography of the volume bias in the media
Sorting candidates by the geography of their constituents (coastal mainland, inland mainland, non-contiguous region, or lacking previous constituents) shows a very interesting pattern.
The ten candidates whose previous constituencies were coastal states, districts, or cities have been covered more. Inland candidates have been covered less. This is actually a remarkably consistent division; it’s also understandable given the geography and demography of the major media outlets.
There are no inland candidates who have unambiguously positive volume bias; the closest are Pete Buttigieg and Amy Klobuchar, who are close to neutral by most measures. There are no coastal candidates with unambiguously negative volume bias; the closest is Joe Biden, who nevertheless leads the coverage race in absolute terms.
Since the political battlegrounds of the Electoral College are neither located in the Acela Corridor nor the West Coast, a geographic bias in the media may impede the nomination of a more electable candidate.
Impacts and effects
From the analysis above, there are a few clear conclusions. First, we can clearly identify a handful of candidates who seem to have attracted a disproportionate amount of media coverage. Elizabeth Warren, Cory Booker, Kirsten Gillibrand, and Bill de Blasio all feature prominently in media coverage compared to candidates who had similar polling numbers. Since Elizabeth Warren’s coverage in “blue” media has also been consistently positive in tone, this has likely fueled her rise.
Second, we can identify two candidates who have very clearly affected by media bias. Both Andrew Yang and Beto O’Rourke have experienced disproportionately sparse and negative media coverage.. It is no wonder that Beto struggled and chose to drop out, or that Andrew Yang’s supporters are griping about the #YangMediaBlackout.
Based on his fundraising and polling, Andrew Yang should have been a larger part of the public conversation earlier. The consequences of failing to do so are negative.
Third, there are a number of lower-profile candidates who were starved for coverage and haven’t been able to get enough air time to make their case to prospective voters. This includes John Hickenlooper and Tim Ryan, who have dropped out; as well as Michael Bennet, Steve Bullock, John Delaney, and Joe Sestak. All of these candidates have backgrounds that are more typical for presidential candidates than Pete Buttigieg.
Fourth, there’s some evidence for bias in volume to support the frequent accusations of bias in tone made by supporters of Bernie Sanders and Tulsi Gabbard. Neither of those candidates, nor Andrew Yang, appear likely to drop out.
The easily demonstrable media bias, if it continues, is likely to cause harm to the eventual Democratic nominee. The DNC may face another round of accusations about trying to rig the primary contest. Media outlets themselves also stand to lose credibility by engaging in blatant bias, either in volume or in tone.
Given that accusations of bias both by the DNC and by “blue” media have been credible and warranted, it’s all the more important to push for fair, transparent, and impartial coverage of the candidates going into the primary.