In this post, we investigate an open-ended text response to the seemingly simple question of what electability means. On our latest survey, we asked respondents,
Lately, there has been a lot of talk about whether candidates for office are "electable" or not. When you hear this being talked about, what do you think people mean by "electability"?
In short, we find that electability is what you want it to be. Voters define “electability” in a highly tautological fashion -- basically, as “the ability to get votes.” The only consistent concept expressed across subgroups includes the ability for a candidate to appeal to “a broad set of voters,” but little beyond that. The term simply does not appear to have much meaning.
Concepts Associated With Electability
In order to examine how voters defined electability, we combined some standard text analysis techniques along with some good old-fashioned reading. Those text analysis techniques included pulling out common terms and short phrases and drawing what we can from their distribution among the population of respondents and among some subgroups of interest. We also provide some representative examples of responses.
For the text analysis side of the project, we began by simplifying responses down to their most basic conceptual components through a few steps. We removed “stop words,” the short, widely used words that do not provide much semantic meaning (“the”, “for”, “its” being examples). Stemming is a common practice in computational text analysis where words are made more comparable by removing parts of the word that change their tense or plural status without fundamentally changing the content of the statement, for example, “electability,” “elected,” and “elects” all become ‘elect’. These stems represent general concepts. We will drop this stemming practice for another text analytic approach focused on the use of adjectives by our respondents.
The following table shows the frequency of the top few stems among all voters. The number one term, used in approximately a quarter of all responses, was ‘elect’. In addition, there are a variety of frequently used stems that reflect general election process concerns: ‘vote,’ ‘win,’ and ‘candid(ate).’
Stems by themselves can be hard to interpret so we also include a variety of what are known as n-grams. N-grams are terms, typically stemmed in practice, that are found next to each other in a response. The second column shows sets of two-term pairs (bigrams), the third column shows sets of three (trigrams) and the fourth sets of four. In each case, the top n-grams indicate that electability is tautologically defined. For bigrams the top two bigrams are ‘elect offic’, and ‘win elect’, the top two trigrams are ‘win general elect’, and ‘enough vote win’, and for four-grams the top two are ‘abl win general elect’ and ‘appeal broad rang voter’. For voters, from conceptualizations ranging from “winning elections” to “enough votes to win,” electability is simply the ability to be elected.
And the following plot shows the frequencies of the top few bigrams in our responses. Among about 1,000 responses, these bigrams showed up just over 200 times in our data. These responses are both purely definitional and ubiquitous among respondents. The phrase “electability” is generally not associated with any specific qualities among many voters.
Beyond that, there are some n-grams that indicate electability means that a candidate is popular or likable. Phrases such as ‘people like’ and ‘appeal enough people’ pop up near the top of the frequency charts. There is not any clear indication beyond that what makes a candidate likable or appealing to voters broadly, though this is a reminder that even if “electability” is not itself a distinct concept from “broadly likeable,” being “broadly likeable” measured by simple approval ratings is an extremely important component of the electoral context. We simply find little evidence that “I approve of this candidate” and “this candidate is electable” are widely divergent concepts.
The story is not much different among various subsets of voters of interest. The following table shows the frequency of stems, bigrams, and trigrams among voters under 45. The same sort of tautological definition appears with an indication that the candidate needs to be popular.
The following table shows some of the top n-grams among Democrats, Independents, and Republicans. The top three terms are the same across all groups. The only difference is that Democrats tend to mention that an electable candidate “can beat Donald Trump.” In contrast, Republicans actually mention New York Representative Alexandria Ocasio-Cortez, although this is used to indicate the ridiculousness of electability. One example: “Alexandria Ocasio-Cortez was not electable or even qualified for any important job.”
Variability of Electable
In an attempt to get a better idea if there were any consistent patterns for defining electability we read through a portion of the responses to see if anything was missed by the automated text analysis. This led to the same conclusion: electability either means ‘can be elected’ or whatever the voter wants. A few examples of the variation for the second type of response are below.
"Sounds like Madison Ave building a model for a politician. How he looks, talks , dresses, has a stepford wife and family. Looks good on tv”
“An emphasis on personality and gender/race/diversity instead of on policy”
“Worthy of holding office because they are committed to improving people’s lives not their own fortune and power.”
“Electability refers to the presence or absence of unpopular actions or decisions in the candidate’s past (skeletons in their closet), and political stances that align with their party and the populace.”
“That's typically a buzz word kicked around by the Leftist media and repeated by Leftist politicians to mean, ‘Does the candidate fit with OUR criteria.’ It has NOTHING to do with what the general public thinks of the candidate. Its 90% propaganda and 10% twisted truths and half-truths.”
Adjectives Associated With Electability
We also pulled out all of the adjectives used by respondents in their open-ended responses in order to understand the descriptive characteristics of electability with which voters are concerned. Democrats overall seemed to be at least partially in on the joke, with the word “white” showing up in their top twenty most common adjectives. Typical responses that contained the word include:
“mainstream, not very old or very young, often white male”
“I think political analysts use electability to say they will appeal to straight, white, christians. It means they aren't too black, too gay, or too liberal. It means they probably won't raise the taxes on the people who are paid to talk about politics.”
It's a stupid designation that usually means "how white, old and male is the candidate"
I think political analysts use electability to say they will appeal to straight, white, christians. It means they aren't too black, too gay, or too liberal. It means they probably won't raise the taxes on the people who are paid to talk about politics.
This is coded language against female and minority candidates that treats differences between the general perceptions of them versus white male candidates as a liability.
The following plot breaks out the use of adjectives by party identification. Once again, phrases like “get enough votes,” “able to win,” etc., indicate a not terribly deep concern with the concept. When asked to report what they think “electability” is, voters -- perhaps unsurprisingly -- are mostly reporting fairly shallow, definitional thoughts.
Who Thinks The Most About Electability?
Finally, we can take a look at who in our sample is thinking the hardest about the question of electability. Using a standard text classification measure of readability, for example, we can compute the approximate “reading level” (we mean that in the same way that the backs of the books you’d get from Scholastic Magazine meant it -- the approximate grade you should be in to be able to read the thing) of each of our responses.
The average response was written at around a 10th grade reading level, which is slightly above the approximate US average of an 8th grade reading level as one would expect from a ssample of registered voters. We only had a handful of respondents offer up a college essay on the subject.
What kind of respondent is thinking the most about this subject? To investigate this question, we can take the complexity of responses and correlate them with some demographic and political factors that might be associated with loquacity on the subject. For example, some might think that more ideological voters think harder about the question of “electability,” or perhaps Democratic partisans think about it more in an era when “defeating Trump” seems to be a major qualifier of candidate electability (and perhaps people with college degrees have more experience answering open-ended essay prompts...).
In the following plot, we show the regression coefficients of a model where the outcome of interest is the approximate grade level of a respondent’s answer to the electability item, and the correlates include respondents’ age, race, ethnicity, education, party identification, sex, and ideological position. Here, we rescaled ideological position so that more extreme values in either direction (“more liberal” and “more conservative” being “higher,” for example) -- although if you run the model without doing this, it makes no difference to the substantive conclusion. We do not see more ideological voters writing more detailed or complex responses to the item, nor do Democrats seem to think any harder about the issue than Republicans or Independents.
Indeed, voters who are older and who have a college degree are the ones more likely to have more detailed thoughts about electability. But being older and being more educated are probably correlated with having more to type on most any given subject, not just electability. Partisans and ideologues think about electability in no greater depth of concern or detail than the rest of us.
Try as we might, we have not been able to imbue the term “electability” with meaning. Voters do not seem to draw much inspiration from the term, and in the minds of voters the term is as ill-defined and uninformative as it is on cable news. Perhaps there are better things to talk about.
John Ray (@johnlray) is Senior Political Analyst at YouGov Blue and a Senior Adviser to Data for Progress.
Kevin Reuning (@KevinReuning) is an Assistant Professor of Political Science at Miami University and a Senior Adviser to Data for Progress
On behalf of Data for Progress, YouGov Blue fielded a survey of US voters nationwide. The survey fielded 3/30/19-4/3/19 on 1,012 US voters. Responses were weighted to be representative of the national population of US voters by age, race, sex, education, and US Census region.