Nixon, Kennedy, and the Victory of Style over Substance

Kennedy was considered the victor by those who watched on TV, while Nixon was thought to be the winner by radio listeners, driving home the importance of visual media in shaping our perceptions.

Even as a non-American, I’ve heard statements to this effect many, many times regarding the first Kennedy v. Nixon presidential debate of 1960. This time, the quote comes from the book Dress Code by Véronique Hyland, a collection of essays about modern fashion, including the role of dress in politics.

But, as always, selection effects are the most powerful force in the universe. People could presumably choose which medium they wanted to listen to the debate in. This means that there could have been a correlation between whether an individual was a Republican or Democrat, and whether they listened to versus watched the presidential debate.

I decided to dig in to this, and then realized that thankfully, someone had already done the work for me. So in the following, I’ll be relying heavily on the 2017 paper Debunking Nixon’s radio victory in the 1960 election: Re-analyzing the historical record and considering currently unexamined polling data by Bruschke and Divine.

Did Nixon even win on the radio?

I only intended to look into the selection effect of what sort of people were likely to listen to a debate versus watch it on television, assuming that the fact of the debate outcomes was well-established.

This was, apparently, a generous assumption: according to the Bruschke and Divine paper, there were only three sources at the time which found that Nixon won over radio. Of these sources, only one of them is anything like a systematic survey¹:

The third account is a survey conducted by Sindlinger and Company that is not included in the company’s published presentation of its own data on the election. A substantially edited and anonymously authored account of the survey appeared in the trade journal Broadcasting. In that account, 178 of 282 radio listeners opted to pick a winner for the debate, and by a better than 2-to-1 margin they picked Nixon. Television viewers in the account picked Kennedy by fewer than 2 percentage points.

Even without selection effects, this is not a lot of evidence on which to pin vast and stubborn theories about politics and sociology—the sample size was less than 300 people! Not to mention that the roughly 37% of people who didn’t pick a winner seem to be edited out of the narrative.

Furthermore, Bruschke and Divine note 30 other samples which collected survey data about the outcomes of the first 1960 presidential debate (though without necessarily recording whether individuals listened to vs. watched the debate). These surveys generally show a much larger margin of victory for Kennedy than the couple of percentage points found by Sindlinger and Company. Some, such as a Gallup poll (#636), have a couple thousand respondents—this poll is particularly interesting for finding that Nixon’s radio gains were only among Republicans. Bruschke and Divine thus suggest that the Sindlinger and Company survey is an outlier.

And even beyond over-reliance on one not-that-large, not-that-representative survey, there are indeed potential selection effects, as Sindlinger and Company didn’t collect party affiliation data.

[Vancil and Pendell] offer plausible reasons for why the radio audience might have been predisposed to vote Republican. In particular, radio listeners were more likely to be from rural areas who lacked television access or west coasters who, due to the time of the debate, listened to it during their workday commute. Both groups were also more likely to be Republican. Indeed, they draw on separate survey data from Sindlinger to show party affiliation was an extremely strong predictor of opinions about the debate; only 4.4% of people who wanted Nixon to win the election thought Kennedy won the debates, and an identical 4.4% of people who wanted Kennedy to win the election thought Nixon won the debates. These figures are especially important, since Vancil and Pendell make simple extrapolations to demonstrate Nixon’s 2-to-1 radio victory margin would have required a massive Democrat Defection effect—something that seems extraordinarily unlikely given the 4.4% finding.

(In this quote, Vancil and Pendell refers to the 1987 paper The mytho of the viewer-listener disagreement in the first Kennedy-Nixon debate by D.L. Vancil and S.D. Pendell.)

So, it seems that perhaps this one individual survey is insufficient for broad, sweeping conclusions about society.

Was there other evidence?

• The American National Election survey of 1960 collected broader data about TV and radio exposure and political partisanship up to the election in that year, though they didn’t collect data specifically about that first debate. They found that “Nixon’s debate victory among the radio audience is almost entirely among Republican listeners.” (Quote by Bruschke and Divine, who present results from the survey in their paper.)
• A 2005 paper by Druckman titled The power of television images: The first Kennedy-Nixon debate revisited randomized 210 students from the University of Minnesota to listen to this debate on the radio versus watch it on TV. Druckman found a statistically significant effect wherein TV viewers were more likely than radio listeners to believe that Kennedy performed better in the debate (means of 2.57 vs. 3.28 on a 7-point scale, where lower numbers favor Kennedy). On the other hand, radio listeners still believe that Kennedy won, and Druckman didn’t collect student partisanship information. Bruschke and Divine also point out that students in Druckman’s sample were (statistically) significantly more likely to answer factual questions about the debate correctly, indicating that perhaps Kennedy’s larger margin of victory could be due to other factors than style triumphing over substance.
• A different study by Pendell and Vancil, this one from 1990 and titled An experimental study of viewer-listener disagreement in the first Kennedy-Nixon debate randomized 550 students to listen to vs. watch the debate. They found that both groups overwhelmingly though Kennedy won (and by similar margins), though “Republican and Independent viewers were significantly less inclined toward Nixon than listeners, and Republican viewers were more undecided after the debate than listeners.” A downside of both this study and Druckman’s is that students listening to this debate after the fact are missing the context of the audience in 1960, and furthermore may already have preconceived notions about this debate due to its cultural standing.

Overall conclusions

• Kennedy probably won the 1960 presidential debate, potentially by a wide margin.
• Political partisanship strongly predicted opinions on debate performance, which I don’t find surprising.
• It’s not even obvious (as per the Bruschke and Divine paper) that Kennedy won on style rather than substance—the two are, after all, not mutually exclusive.

And finally, a very prescient and of-the-moment point that Bruschke and Divine make, and which I was unfamiliar with, was that Nixon was ill at the time of the first debate. Describing media coverage at the time, they write

Nixon is not described as unattractive in the first debate; he is described as sick… The health and vitality of global leaders can almost be considered substantive issues.

Footnotes