The first televised debate between US President Donald Trump and Democratic challenger Joe Biden will take place on 29 September. The US media often portrays this as a make-or-break moment in the campaign. One CNN analysis claims that “[t]he best opportunity [Trump] will have to alter the race's trajectory is the three debates with Biden, especially the first one, on September 29, since traditionally that is the one that draws the most viewers,” and Vanity Fair magazine says that “[t]he presidential debates of 2020 will likely be this country’s most dramatic and consequential political theater of the 21st century.”
But how many voters will change their choice based on what they see?
A large number of US-based studies examine this question. While some conclude that TV debates truly matter (e.g. Hillygus and Jackman 2003), others find modest or even null effects on vote choice (e.g. Shaw 1999). However, these studies generally focus on a unique election or a small number of races in the same country, and so their estimates may capture the effects of concomitant events and underlying time trends.
Another set of studies use rigorous experimentation in developing countries and find that debates can build voter knowledge and affect voter choice (Brierley et al. 2019, Bidwell et al. 2019, Platas and Raffler 2019, Bowles and Larreguy 2020). Brierley et al. (2019) show debates also reduce polarisation in Ghana, while for Sierra Leone Bidwell et al. (2019) trace the effects all the way through to the behaviour of elected members of parliament, and give evidence that debates increase accountability and lead to policy change.
However, these studies are quite far removed from the US electoral season, focusing instead on debates in low-income democracies with scarce information. The debates feature parliamentary candidates (not executives) and they are not televised but featured on smartphones, in public gatherings, or on the radio.
In a new paper, two of us – Caroline and Vincent – use survey data from 62 elections in ten OECD countries to begin to fill this knowledge gap (Le Pennec and Pons 2020). Voters are surveyed not once, but twice – before and after election day – which provides new detail on how and when they make their choices.
We find that across the sample of countries, a significant portion of survey respondents can still be swayed in the last 60 days in a campaign – enough to change the outcomes of many elections. However, this share is far lower in the US, and across all countries and elections considered, TV debates do not substantially impact vote choice.
This suggests a campaign wishing to sway last-minute voters might do better by focusing on individual outreach – a challenging prospect, given the Covid-19 pandemic.
In the paper, we seek to identify the overall impact of information voters receive across an electoral season, and how that compares to other factors such as party affiliation in determining their vote. We assemble a dataset of nationally representative surveys conducted around 62 elections in ten countries from 1952 to 2017, resulting in a total of 253,000 observations.
These surveys entailed interviewing a new set of people every day before the election to elicit their voting intentions, then interviewing them again after the election to record their actual vote choice. Comparing voters’ responses to the two surveys reveals whether they had actually settled on their final vote choice by the time of the pre-electoral survey, removing recollection bias which can arise when asking voters when they made their choice and what made them choose.
The paper sheds light on a number of determinates of voter choice, such as policy preferences (information received during the electoral season does not appear to affect policy preferences expressed before and after an election), candidate electability (surveys suggest that voters’ beliefs on candidates’ attributes and policy positions matter more than beliefs on their chances of victory), and the perceived importance of different issues (which changes during the campaign and correlates with shifts in vote choices).
But three of the findings speak directly to the possible effects of the 29 September debate:
1. Across the sample, sufficient numbers of voters make their choice in the final two months to sway many elections
As Figure 1 shows, the surveys reveal that the fraction of people with identical pre-election vote intention and post-election vote declaration increases by 17 percentage points over the last 60 days before the election, from a baseline of 71%. That increase comes both from respondents becoming more willing to state any vote intention, and from that statement becoming more consistent with their final reported vote.
Figure 1 Vote choice consistency as a function of the number of days relative to the election
2. Far fewer voters are subject to last-minute changes in the US than other countries
The influence of campaigns on vote choice has been relatively stable for the last 70 years but it varies substantially from one country to another.
Figure 2 represents this, by showing vote choice consistency across countries and over time. For each election, we estimate the average daily increase in the fraction of people with identical vote intention and vote declaration in the last 60 days before the election. The figure plots these election-specific estimates – the daily increases in vote choice consistency – against election year, along with country-level linear fits. The share of voters who change their mind in the final two months of an election season is much lower in the US than in other countries and this has changed little over the decades.
Figure 2 Average daily increase in vote choice consistency across countries and over time
3. TV debates change few minds
The study sheds light on the effects of 56 TV debates in 31 elections in seven countries, which not only gives the results the advantage of statistical power, but also a broader focus than studies such as Hillygus and Jackman (2003) and Shaw (1999), which focus on a few elections in a single country.
The methodology confers two further advantages in identifying the effects of debates. First, the fact that there are many debates in the samples decreases the risk that the event study captures the effect of other shocks affecting vote choice – as we can reasonably assume such shocks should be uncorrelated with the dates of debates. And second, debates took place at different times in different elections, which allows an ability to control flexibly for the consistency of vote choice that rises non-linearly in the 60 days leading to the election, as we saw in Figure 1. In other words, effects of debates can be disentangled from those underlying time trends.
The results show no significant impact of TV debates on individual consistency between vote intentions and vote choices expressed before and after an election – or any of the other relevant factors considered, such as voter policy preferences, issue salience, or beliefs on candidates. Thanks to the precision of the estimates, we can reject any impact higher than 0.5 percentage points on vote choice consistency, on average. Furthermore, debates do not contribute to vote choice formation for any type of voters and in any context, including close and uncertain races.
If not debates, what will sway last-minute voters?
Unlike televised debates, door-to-door canvassing by political activists has a strong potential to mobilise and persuade voters – at least in France, according to a series of studies one of us (Vincent) conducted with co-authors. It can bring in new voters through registration (Braconnier et al. 2017) and mobilise a party’s existing supporters to vote (Pons and Liegey 2019).
Vincent also gave evidence that canvassing can perform that function so often attributed to TV debates – changing minds. He embedded a countrywide experiment in François Hollande’s campaign in the 2012 French presidential election (Pons 2018), randomising who received visits as an estimated 80,000 left-wing activists knocked on 5 million doors to encourage people to vote for Hollande. The study found that canvassers did not affect turnout, but increased Hollande’s vote share by 3.2 percentage points in the first round and 2.8 points in the second, accounting for a quarter of his victory margin.
What bearing will these findings have on an election in the United States, and especially one taking place during a global pandemic? Republicans and Democrats are taking a dramatically different approach to field work. Donald Trump’s campaign recently claimed to have knocked on 1 million doors in one week, while Joe Biden’s campaign knocked on zero, choosing instead to focus on safer methods such as phone calls, texts, and virtual organising.
It is hard to say if such efforts will change many voters’ minds, but if past evidence holds true, the TV debates will not.
Bidwell, K, K Casey, K and R Glennerster (2020), “Debates: Voting and Expenditure Responses to Political Communication”, Journal of Political Economy 128(8): 2880-2924.
Bowles, J and H Larreguy (2020), “Who Debates, Who Wins? At-Scale Experimental Evidence on Debate Participation in a Liberian Election”, Working Paper.
Braconnier, C, J Y Dormagen and V Pons (2017), “Voter Registration Costs and Disenfranchisement: Experimental Evidence from France”, American Political Science Review 111(3): 584-604.
Brierley, S, E Kramon and G K Ofosu (2019), “The Moderating Effect of Debates on Political Attitudes”, American Journal of Political Science 64(1): 19-37.
Hillygus, D S and T G Shields (2009), The Persuadable Voter: Wedge Issues in Presidential Campaigns, Princeton University Press.
Le Pennec, C and V Pons (2020), “How Do Campaigns Shape Vote Choice? Multi-Country Evidence from 62 Elections and 56 TV Debates”, NBER Working Paper 26572.
Platas, M R and P Raffler (2019), “Meet the Candidates: Field Experimental Evidence on Learning from Politician Debates in Uganda”, in Dunning et al. (eds.) (2019), Information, Accountability, and Cumulative Learning: Lessons from Metaketa I. Cambridge University Press.
Pons, V (2018), “Will a Five-Minute Discussion Change Your Mind? A Countrywide Experiment on Voter Choice in France”, American Economic Review 108(6): 1322-1363.
Pons, V and G Liegey (2019), “Increasing the Electoral Participation of Immigrants: Experimental Evidence from France”, The Economic Journal 129(617): 481-508.
Shaw, D R (1999), “The Impact of News Media Favorability and Candidate Events in Presidential Campaigns”, Political Communication 16(2): 183-202.