Media’s Voting Record

Election 2016           Social Desirability Bias / Media Bias:  Media’s Voting Record

As explained elsewhere, the national media play a central role in setting the tone for political discourse.  Certainly, it is the transference mechanism that can most effectively show the general public which parties, candidates, and policies should be considered socially acceptable.  In this sense, the media also play a central role in determining the existence of Social Desirability Bias in national politics.

If the media holds opinions on a particular party, candidate or policy that are significantly worse than those of the general public, it likely will produce Social Desirability Bias which will make polls look worse for the ‘socially undesirable’ party, candidate, or policy, but not necessarily make the election outcome worse.  In fact, in theory, the socially undesirable candidate will gain votes ‘unexpectedly’ on election-day as supporters vote anonymously.

One way of measuring bias of the media is to review their voting record.  Given that the office of the president has more or less gone back and forth between Democrat and Republican during the last century, we would expect that the voting record of the media might reflect these trends.  In other words, we might assume that over long periods of time, roughly half of the journalists’ votes went to each major party.

Journalists, it seems, vote almost exclusively against Republicans.  Furthermore, this trend has been intact presumably for decades.  There are many caveats, which will be explained later, associated with these numbers so please read the entire post.  The results are summarized in the following table.

Table 1:  Percent of Journalists Having Voted or Declaring Intention to Vote for the Republican Candidate in US Presidential Elections 1964 – 2016 

Source:  The Argus-Press, Feeding the Beast: White House versus the Press, Slate 2012, Slate 2000 – 2008, California State University at Los Angeles, Freedom Forum poll of Washington bureau chiefs and congressional correspondents, University of Connecticut Department of Public Policy, The Media Elite,  Los Angeles Times

There are many things that jump out from this data but certainly the most obvious is that these numbers are unusually low.  From 1964 to 2012, Republicans held the White House slightly over half the period.  It is therefore surprising that the maximum support from journalists for a Republican candidate only reached 26%, and that year was the landslide reelection of Reagan who was one of the most popular presidents, Democrat or Republican, in recent memory.

The first column shows surveys conducted on the general media which includes mainstream TV and news publications.  This shows that going as far back to the 1960s there was an obvious skew toward the Democrat and away from the Republican candidate.  The 2016 data is taken from Politico and shows that the general historical trend appears to be intact.  In fact, if this expected 2016 figure turns out to be correct, it shows that the general media is at its least supportive moment for a Republican candidate since 1964 when Goldwater (Republican) lost the election in one of the worst Electoral College losses on record (52 to 486).  Interestingly, there are some similarities between Goldwater and Trump as both challenged established Republican policies and upset many of the party leaders to such an extent that they would not campaign for them.  Regardless of similarities between candidates, this election year appears to have an unusually unsupportive media for Trump given voting intentions.

The White House Press Corps, the second column, similarly shows very low support for Republican candidates.   Although there is little overlap with the first column, they appear to be more or less in-line with each other.  The Press Corps is usually stationed at the White House and covers the executive branch, events at the White House, press conferences, etc.  The 2016 data point of zero is based on various reports that none of the White House Press Corps is a Republican.  This seems very unusual but is not unprecedented as a 0% figure also appeared in 1984.

Assuming these figures are correct Trump appears to be setting the pace for the lowest media support on record for a candidate of a major party.

The remainder of the data is used to confirm the general trends and levels of support.  In general, we see that various data, though spotty, depict the media as unsupportive to Republicans.  One potential highlight is that newspaper staff appears to be slightly more neutral.

Now for the caveats, there are many.  This data is not that great.  Let’s get that out of the way. 

Many of the problems inherent in this data appear in other datasets as well.  Business and especially financial markets are forced to deal with non-perfect data every day.  In those cases, many times a weight of evidence approach is used in that no single data point or dataset is conclusive or seen as high-quality but given that all point in the same direction you can infer enough of the situation to make an accurate conclusion.  With this data, we appear to be in this situation.  By themselves, none of these datasets would be enough to really make a strong conclusion, but given that they all point to journalists voting Republican very rarely and given that indications point to Trump being even more unpopular with journalists we can infer a strong conclusion.

This last point of Trump being even more unpopular with journalists is backed up by superficial examples of Trump repeatedly complaining about the media, much more so than any candidate in recent memory, and the general media referring to him in extremely derogatory terms.  It does not seem like the conclusion that Trump is more unpopular with the press than previous candidates or that the media in general prefer Democratic candidates should be surprising.  The real shocker is the size of the support for the Democrats over long periods of time.

Getting back to the issue of data quality, it should be noted that much of the data was compiled by Republican or conservative sources and/or found on conservative leaning sites.  This is a red flag for neutral analysis, however, the data is fairly well footnoted.  For instance, the book The Media Elite was used for the data in the first column from 1964 to 1976.  This book was one of the first studies in liberal media bias and has a definitively pointed conclusion.  Additionally, many of the data points were taken from a report entitled Media Bias 101 which again seems to have a rather strong conclusion from the start.

Having noted the potentially biased sources (discussing bias which is somewhat ironic), it should be pointed out that the sources cited in many of these reports and sites are from non-conservative sources.  For instance, various university and newspaper surveys are referenced.  In general, the data appears convincing but it would be much more so if there were a counter argument that could be analyzed as well.  Such a counter argument might exist (that the media and journalists do not vote consistently for Democrats) but we could not find it.  In other words, we could not find reports, books or surveys to refute any of the claims.  In the absence of a counter argument or even data showing these surveys are not accurate, as an analyst you have to side with the well-footnoted data coming from a variety of sources.

The best arguments against these findings are that the data was not collected in a scientific manner and the sample sizes were low, extremely low in some cases.  Most of these studies were conducted seemingly out of curiosity.  For example, a journalist sending out surveys to his/her colleagues or a newspaper contacting journalists and asking them which candidate they supported are great for casual observations or a news story but not for in depth studies.

The source for both data points from 2016 can be described as surveys done more for a news story than for real analysis.  Politico created an infographic around a survey of 82 journalists.  This number is rather low, but gets even lower for those who responded to who they will vote for in the election (Trump received 1 vote and Clinton 12 votes).  Though extremely low, this figure is supported by other points within the survey such as Clinton winning the question of “who do you think will be the next president?” with 86% to Trump’s 10%.  Additionally, 44 of 44 respondents answered the question of “which news organization is doing the best job covering the 2016 campaign?” with a liberal leaning response – with 75% of the respondents answering either The New York Times or The Washington Post, generally accepted as two of the most liberal leaning newspapers.  So, yes, the sample size is very low and it could distort the data.  However, various data within this survey point to the same conclusion that journalists in general favor Clinton and/or are liberal-leaning to a lopsided extent.

The other 2016 data point has a problem as well.  It is based on Politico survey that also suffers from a low sample size and that appears to have been conducted in a rather casual manner.

Another questionable addition to the table is the Slate data.  Slate is clearly a left leaning news agency so comparing the results of the percent of its staff intending to support Trump could be interpreted as disingenuous.  The reason for including this data is for comparison purposes.  For instance, you might assume that the figures for Slate’s staff would not be in the same range as for the rest of the media knowing its strong and obvious liberal skew.  However, looking at the totality of the data, the Slate responses seem lower but not terribly so.  In short, the Slate data help to put the other data points in context as to where the media are on the political spectrum.

In conclusion, the media certainly appears to have a bias against Republicans from at least the 1960s given their apparent pro-Democratic voting record and Trump, using very limited data with low sample sizes, seems to be one of least supported candidates in terms of media voting intentions.  Though there are many weaknesses to the data, the fact that they all confirm one another helps to strengthen the conclusion.