do presidential debates matter? evidence from a central and … · 2017-08-31 · keywords: impact...

33
Do presidential debates matter? Evidence from a Central and Eastern European Context Costin Ciobanu McGill University August 31, 2017 Abstract This paper analyses the effects of the 2014 presidential debates on vote intentions in Romania. Although there are good theoretical reasons to expect presidential debates to matter in Central and Eastern European countries, the paper demonstrates that presidential debates only have a significant effect on the second-round vote choice of first-round thwarted voters and first-round non-voters. Exposure to the debates also increases the probability that a voter will form a second-round vote intention. The debates have little impact on those who had voted in the first round for the candidates who qualified for the run-off; for this subset of voters, the role of the debates seems to be one of reinforcing previous beliefs and behaviours. On the other hand, there is strong support for the thesis that voters are more likely to say they will vote for the candidate whom they perceive to have won the presidential debates. The paper calls for more research on the role of presidential debates to be conducted in Central and Eastern European democracies (comparatively or in different electoral years) and in countries with two-ballot electoral systems, which experience different dynamics than the United States. Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting behaviour, Media effects, Two-round elections.

Upload: others

Post on 12-Jul-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Do presidential debates matter? Evidence from aCentral and Eastern European Context

Costin CiobanuMcGill University

August 31, 2017

Abstract

This paper analyses the effects of the 2014 presidential debates on vote intentions in

Romania. Although there are good theoretical reasons to expect presidential debates

to matter in Central and Eastern European countries, the paper demonstrates that

presidential debates only have a significant effect on the second-round vote choice of

first-round thwarted voters and first-round non-voters. Exposure to the debates also

increases the probability that a voter will form a second-round vote intention. The

debates have little impact on those who had voted in the first round for the candidates

who qualified for the run-off; for this subset of voters, the role of the debates seems to be

one of reinforcing previous beliefs and behaviours. On the other hand, there is strong

support for the thesis that voters are more likely to say they will vote for the candidate

whom they perceive to have won the presidential debates. The paper calls for more

research on the role of presidential debates to be conducted in Central and Eastern

European democracies (comparatively or in different electoral years) and in countries

with two-ballot electoral systems, which experience different dynamics than the United

States.

Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential

elections, Voting behaviour, Media effects, Two-round elections.

Page 2: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

1 Introduction

Do presidential debates matter for the outcome of elections? Do they influence vote

choice or have they simply become a new form of political entertainment with little electoral

impact? Given the lack of research, answers to these questions cannot be provided for

young democracies, such as those of Central and Eastern Europe. The situation is different

for mature democracies, where these questions have been studied since the first Kennedy -

Nixon presidential debate (Chaffee, 1978; Hall – Jamieson et al., 1988; Schrott, 1990; Blais

& Boyer, 1996; Schroeder, 2008; Gerstle, 2008).

Although there have been instances where the presidential debates have impacted the

outcome of the election (for example, the 2000 US election), the overall consensus is that

presidential debates have a weak influence on vote choice (Holbrook, 1996). This is generally

due to the importance of longer-term structural variables over short-term campaign factors

(Schrott & Lanoue, 2013) as well as the methodological difficulties of estimating debate

effects (Erikson & Wlezien, 2012). To address the lack of knowledge of debate effects in

young democracies, this paper analyses the impact of the two televised presidential debates

that took place between the first and second rounds of the 2014 Romanian presidential

election. The paper draws on survey data to examine whether exposure to the debate and

perceptions of who won had a significant effect on vote intentions and, more generally, on

the outcome of the election.

Typical of young democracies in Central and Eastern Europe, the Romanian electoral

context is characterized by weak ideological identities and partisan attachments, with few

voters having a deeply socialized party identification. This means that voters’ evaluations

of the candidates’ debate performances are less likely to be biased and consequently debate

effects could be expected to be stronger than those observed in the United States and other

established post-industrial democracies. This seems especially likely given the large number

of undecided voters and the fact that the second of the two debates took place only four

days before the second-round vote and would thus not be subject to decay. Moreover, the

2

Page 3: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

two-ballot electoral system, adopted by Romania and other countries in the region, creates

dynamics that do not exist in most major Western democracies (such as the US, Germany

and Canada). The existence of thwarted voters and first-round non-voters increases the

potential for presidential debates to be more influential in these new democracies. The

paper will explore whether these expectations are confirmed by empirical tests.

I start by presenting the literature on the impact of presidential debates, highlighting

the lack of research on new democracies, such as those of Central and Eastern Europe.

After introducing the hypotheses to be tested, I describe the context of the 2014 Romanian

presidential elections. The next section describes the data and measures as well as the model

specifications. After presenting the findings, the paper concludes with a discussion of the

implications and future research directions.

2 Assessing debate effects

Scholars in North America and Western Europe have devoted a good deal of attention

to the effects of face-to-face televised debates between contenders for top political office.

As McKinney (2008) notes, presidential debates are viewed as “one of the most important

forms of campaign communication by the public, political candidates, and the media, as well

as by political communication scholars” (p. 59). McKinney and Carlin (2004) distinguish

four types of presidential debate effects: behavioural, cognitive, latent, and candidate image

evaluation effects. Behavioural effects refer to whether exposure to the debate influences a

citizen’s vote choice. An important distinction in this respect, with important methodological

consequences (Prior, 2012), is made between direct and indirect behavioural effects: the

former refer to those who watched the debates and the latter to those who were exposed

(through media coverage or personal networks, for example) to discussions about the debates.

Cognitive effects are related to the idea that the debates facilitate viewers’ acquisition of issue

knowledge by familiarizing them with the candidates’ positions. When it comes to candidate

3

Page 4: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

image evaluation effects, scholars are interested in discovering whether “debate exposure

influences viewer perceptions of candidate character or image traits.” Finally, latent effects

occur when exposure to the candidates’ performance in the debates activates the “citizens’

various civic and democratic tendencies” (McKinney, 2008, p. 164).

Debate effects are dependent on the specific dynamics of the campaign and on the political

profile of debate viewers. Chaffee (1978, p. 342) identifies four conditions for the debates to

have an effect on vote choice: “(a) at least one of the candidates is not well known, (b) many

voters are undecided, (c) the contest appears to be a close one, and (d) party allegiances are

weak”. However, further research has shown that reality is more complex: “debates change

public opinion to the extent that one candidate’s performance is so clearly superior that

even many of those who oppose him must acknowledge the reality of his victory” (Schrott

& Lanoue, 2013, p. 691).

Based on their analysis of American presidential debates, McKinney & Warner (2013,

p. 256) concluded that, “campaign debates do indeed matter, and our analysis of ‘the

trans-campaign effects on such matters as voting behaviour, image formation, and attitude

change’ has allowed us to understand just how debate effects function in several important

ways”. However, the general conclusion with respect to behavioural effects is that, at least

in the American context, “little change in voting intentions is typically recorded following

exposure to debates” (McKinney, 2008, p. 164). Analysing studies conducted on presidential

debates, Schroeder (2008) concludes that presidential debates change perceptions (related,

for example, to issue knowledge, the character traits of the candidates, or the stakes of the

election) rather than voting intentions; in the end, despite their audience, the debates are

only one of the potential factors influencing the vote.

The limited impact reflects the importance of longer-term structural variables over cam-

paign factors (Schrott & Lanoue, 2013) as well as the methodological difficulties of estimating

debate effects (Erikson & Wlezien, 2012). Holbrook (1996, p. 114) argues that campaign

events have a weaker influence than longer-term factors: “the perception of most viewers are

4

Page 5: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

coloured by their political predispositions going into the debate...the single best predictor

of which candidate a viewer thought won a given debate is that viewer’s pre-debate vote

choice”.

Assessing behavioural effects is as complicated as assessing campaign effects in general.

According to Wlezien (2010), two main problems arise. On the one hand, “the effects of

most campaign events are very small” (p. 110). On the other hand, “our ability to detect

campaign effects is limited” (p. 110). Based on their study of US presidential elections from

1952 to 2008, Erikson & Wlezien (2012) conclude that, “it is very difficult to estimate the

effects of some seemingly big events, such as presidential debates. We can measure voter

preferences in trial-heat polls before a debate and compare it to those reported in polls a

few days after...however, much ‘change’ we observe in surveys is simply the natural result of

sampling error...Rarely can we point to a campaign event and say that it made a difference

of a size and duration that can be identified in the polls” (p. 12). Schroeder (2008) goes

further, arguing that it is practically impossible to isolate the debates’ effects on vote choice

from other influences.

If debates are understood as a particular type of campaign event, assessing their impact

on vote intentions is equivalent to determining whether they are part of a stationary series

(the decay of effects is observed) or of an integrated series (the accumulation of effects is

demonstrated) (Wlezien & Erikson, 2000). The difficulty in evaluating their impact is only

increased by the fact that, to use Wlezien & Erikson’s (2000) terminology, some events may

cause a “bounce” (a stationary series) and others a “bump” (an integrated series), while yet

other events may cause preferences to move and then bounce back, but to a level different

from before. This aspect highlights the difficulty of capturing the indirect impact of the

debates and demonstrates the importance of contextual factors – the longer the temporal

distance between the vote and the event, the harder it is to identify its contribution to the

outcome. All in all, without proper attention to these issues, this complex environment

undermines even weak causality claims about the impact of presidential debates.

5

Page 6: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Despite these serious limitations, various methodologies have been used to analyse whether

exposure to debates and evaluations of the candidates’ performance affect vote choice. What

is important to highlight is how dependent on the institutional context these methodologies

are. It matters a lot whether the debate is between party leaders, presidential candidates,

or candidates for Chancellor.

Using individual-level data, Blais & Boyer (1996) investigate whether the debates between

the party leaders influenced the outcome of the 1988 Canadian federal election. They propose

four research designs, each of which poses some methodological difficulties. The first is a

cross-sectional group comparison that looks at the voting behaviour of those who did and

those who did not see the debates. The ex-post cross-sectional group comparison can, with

the appropriate controls, indicate the net direct impact of the debates; however, as the

authors acknowledge, in the presence of potential substantial indirect effects, the approach

will underestimate the real total impact. The second design is a panel analysis, in which

the vote intentions of debate watchers and non-watchers are compared before and after

the debates. Like the previous design, this research design cannot deal appropriately with

the issue of indirect effects. The third approach is a panel reactions analysis, with the

debates being deemed to have an impact if support for the party increased among those

who considered that the party leader had won the debate and decreased among those who

thought that a different leader had won. Although it has the advantage of capturing indirect

effects, the design is vulnerable to the question of whether the debate’s impact on the vote

is entirely channelled through perceptions about the debate winner (Blais & Boyer, 1996, p.

147). The fourth design is a time-series analysis, which consists of looking at the movement of

vote intentions during the campaign and determining if the debates affected that movement.

Even though any shift in voting intentions during the campaign can be influenced by other

events, this research design is considered the most logical and “the most appropriate design

to tap the impact of debates” (Blais & Boyer, 1996, p. 147). The authors highlight how

important the research design is in assessing the behavioural impact of the debates: the first

6

Page 7: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

two designs seem to indicate no debate impact, while the panel reaction and the time-series

analyses demonstrate that the debates had a “substantial and enduring impact on the vote

and that they were decisive in the contest for second place between the Liberals and the

NDP” (p. 143).

Lanoue (1991) also investigates the impact of party leader debates on voting behaviour.

Using three probit models, he estimates the likelihood of voting for each of the three federal

parties in the 1984 Canadian election based on exposure to the debate, while controlling

for factors such as age, education, income, gender, and party identification. The study

demonstrates the importance of appropriate controls in isolating the impact of the leadership

debates. With respect to the impact of the debates, the paper goes against the traditional

idea that national debates mostly reinforce the viewers’ prior preferences; it argues that

the debates had a significant effect on voting behaviour, with the impact being especially

strong among French-speaking Canadians. The debates favoured the candidate most fluent

in French.

Schrott (1990) uses pre- and post-debate panel data to examine whether the debates that

took place between the party leaders in three West German federal elections (1972, 1976,

and 1983) had an effect on vote choice. Like Lanoue, Schrott estimates a probit model,

with individual vote choice as the dependent variable. Two models are compared: the first

model has party identification and pre-debate evaluations of party leaders as the independent

variables while the second model adds evaluations of the debate winner. The conclusion is

that the latter variables increase the goodness-of-fit of the model.

In their aggregate-level analysis1 of 26 U.S. presidential debates, Schrott & Lanoue (2013)

assess the impact of debate performance (as measured by the percentage of those indicating

the candidate as winner) by first establishing a baseline expectation of success. Their starting

assumption is that “many of the factors that influence viewers’ assessments of winners and

losers are exogenous to the content of the debates themselves” (p. 685). Their OLS model,

1 The authors look at the percentage of Gallup respondents naming a candidate as the debate winner in26 US presidential debates.

7

Page 8: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

which incorporates only long-term factors (such as pre-existing candidate preferences, party

identification, and attitudes toward the incumbent administration), shows that “pre-existing

factors having nothing at all to do with the actual performance of the candidates make up

more than 50% of the measure that we typically use to identify the winners and losers of

presidential debates” (p. 687). Like Lanoue (1991), Schrott and Lanoue emphasise the role

of exogenous factors and the need to control for them in order to tease out the impact of

debates.

The typical conclusion of these North American and Western European studies is that

the impact of debates on vote intentions is minimal. However, given the lack of research,

it is not known whether the same conclusion holds for new democracies. Theoretically, we

could expect stronger debate effects than have been observed in established post-industrial

democracies.

First, the low degree of party system institutionalisation in Central and Eastern Europe,

which is evident in greater electoral volatility, non-programmatic and non-ideological linkages

between voters and parties, and personalistic voting (Mainwaring & Torcal, 2006), implies

that party identification will not be as important as in mature democracies and will not serve

as a “perceptual screen” (Campbell et. al, 1960), when evaluating debate performance. Since

the roots of political parties are not deep, one would expect that the effect of the debates

on voters would not be contaminated by such partisan influences. Second, the presence of a

high number of undecided voters at the end of campaigns increases the likelihood that the

debates will matter (this was established by Chaffee (1978) in the case of the US presidential

debates).

Third, the institutional context might enhance the impact of the debates. For example,

two-ballot elections are the norm in Central and Eastern European presidential elections

and the debates usually take place during the two weeks between the first round and the

run-off. Consequently, the existence of first-round non-voters and thwarted voters who need

to choose a second-round candidate opens up the possibility that debates are decisive for

8

Page 9: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

the fate of presidential candidates (the traditional media frame). In the Unites States,

Canada and Germany, countries where extensive research has been conducted on the role

of televised debates, the national elections have only one round and so the potential impact

on thwarted voters and first-round non-voters has been overlooked. Given the widespread

use of two-ballot elections in Romania and other Central and Eastern European countries,

it is important to analyse the impact of debates on these categories of voters. This major

institutional difference could explain why American, Canadian and German findings may

not hold under different electoral rules. This paper will pay particular attention to the

influence of exposure to debates on the voting behaviour of thwarted voters, first-round

non-voters, and undecided (compared to decided) voters. It will then be possible to assess

how generalizable findings about the weak effects of televised debates are in representative

democracies.

Based on these theoretical expectations, we can hypothesize that:

H1: Voters who are exposed to presidential debates will be more likely to change their vote

intentions.

H2: Exposure to presidential debates will have a significant impact on the vote intentions of

thwarted voters and first-round non-voters.

H3: Exposure to the debates increases the likelihood that a voter will form a second-round

vote intention.

H4: Voters are more likely to say they will vote for the candidate whom they perceive to have

won the presidential debates.

3 The 2014 Romanian presidential election

These hypotheses will be tested using the case of the debates in the 2014 Romanian

presidential election. This is the first study in the region to look at the behavioural effects of

presidential debates. The results have the potential to generalize to other countries in Central

9

Page 10: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

and Eastern Europe. In addition to the fact that these countries share fundamental political

similarities with Romania in their transitions from Communism to democracy (Grzymala-

Busse, 2007), many of them also have presidents directly elected in two-ballot systems, in

which the run-off takes place two weeks after the first round (this is the case, for example,

in Poland, the Czech Republic, Slovakia, Croatia, and Bulgaria, as well as Romania). The

constitutional role and symbolic prestige enjoyed by presidents in the region transforms

the presidential elections into first-order elections. Moreover, as in Romania, the debates

typically take place only a couple of days before the final vote, which minimizes the potential

decay of effects. Thus, the Romanian case can be taken as typical of the region in terms of

the likely behavioural effects of presidential debates.

The Romanian electoral system for the presidential elections is very similar to the French

one. If no candidate wins the majority of the votes in the first round, a second round is

organised on the second Sunday following the first vote. The 2014 presidential elections

were necessarily “change elections” since the right-wing incumbent President Basescu was

not allowed to run due to constitutional provisions that limit the number of terms to two2. On

2 November, Romanians went to polls to elect their new President. The main favourites to

become the Eastern European country’s fourth head of state were incumbent Prime Minister

Victor Ponta, leader of the left-wing Social-Democratic Party (PSD), and Klaus Iohannis,

one of the leaders of the right-wing alliance between the National-Liberal Party (PNL) and

the Liberal-Democratic Party (PDL). In the context of a 53.17% turnout (which included

Romanians residing abroad), Victor Ponta obtained 40.44% of the first-round votes, while his

main opponent got 30.37%; no other candidates obtained more than 6%. Two weeks later,

on November 16th, benefiting from a surge in electoral participation3 (64.1% turnout), Klaus

2 Gherghina (2015) provides a detailed account of the electoral context in terms of the electoral system,the candidates, manifestos, and the development of the campaign.

3 The government’s poor organisation of the vote in the countries with large Romanian communities(Spain, Italy, France, Germany, and United Kingdom) captured the attention of the media on November2nd. As more than 3 million Romanians work and live abroad, people had to queue for hours before voting,because of the small number of available polling stations and other bureaucratic requirements. Only 161,054Romanians from the Diaspora voted in the first round, leaving tens of thousands unable to have their say.This had political and electoral consequences: The Foreign Affairs Minister resigned, the Prime Minister

10

Page 11: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 1: Exposure to the two presidential debates

Presidential debateAudience share

(Total active TV owners, ARMA)Exposure to the debate

(Total Population, Sociopol Survey)

First debate 33.1% 63.5%Second debate 20.9% 53%

Iohannis caused a huge upset and won the elections with 54.43% of the votes, becoming the

first elected president of Romania belonging to a national minority.

Between the first and second rounds, two face-to-face presidential debates were organised.

The Realitatea TV news channel hosted the first debate on 11 November (five days before

the second-round vote and three days before the official end of the campaign) and the B1TV

news channel hosted the second debate the next day on 12 November (four days before

the run-off and two days before the official end of the campaign). The debates attracted a

considerable number of viewers, according to both audience studies and surveys (see Table

1).

Figure 1 compares vote intentions for the two candidates who qualified for the run-

off based on four Sociopol surveys conducted between the two rounds (see Appendix A

for information about the methodology of the polls). It is clear that there was a good

deal of variation in vote intentions for the two candidates, Ponta and Iohannis, and a lot

of uncertainty. What is more interesting is that the 95% confidence intervals associated

with vote intentions overlap in both the second and fourth surveys, which were conducted

immediately before and after the debates. This was clearly a very tight race in which the

presidential debates, with their high audience, had the potential to decisively impact the

result.

pledged to better organise the vote (the turnout of Romanians residing outside the country in the secondround was 378,811), and the right-wing candidate used the debacle as his main line of attack against Ponta.

11

Page 12: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Figure 1: The evolution of vote intentions (with 95% confidence intervals) between the firstand second rounds of the 2014 presidential elections (5 – 14 November, Sociopol surveys)

4 Data and Measures

The hypotheses will be tested using data from the survey conducted by the Romanian

polling institute Sociopol in the two days following the second presidential debate (13 – 14

November 2014). The survey sample size was 833 and the response rate 25%4; the method

of data collection was Computer-Assisted Telephone Interviewing (CATI). The sample was

weighted based on the 2012 Statistical Yearbook to be representative of the adult population

of Romania in terms of sex, age, ethnicity, region, and city size. More technical information

about the survey is available in Appendix A.

The dependent variable is based on the answers to two questions. The first is a typical

vote choice question: “Victor Ponta and Klaus Iohannis qualified for the second round of the

4 Per the Sociopol experts, this represents the total response rate; it includes not only the persons whorefused to answer, but also the unallocated phone numbers, the respondents who were not home when thesurvey participation call was made, and the persons who were not considered when the weighting took place.

12

Page 13: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Romanian presidential elections. Which of these two candidates would you vote for?” Only

68.8% of the respondents expressed a vote intention for one of the two candidates, reflecting

the high number of undecided voters only two days before the vote (for descriptive statistics,

see Appendix B). To increase the number of those who had a vote intention, I employ the

answers to a follow-up question for those who answered that they would not vote or that

they did not know or had not answered (N=260), asking: “In any case, which of the two do

you consider the best to be the next President of Romania: Ponta or Iohannis? Including

these leaning voters, the percentage of respondents with a vote intention increases to 78.9%

of the sample (N=657). 51.8% expressed a vote intention for Ponta and 48.2% for Iohannis,

underlining again the closeness of the race. To control for possible differences between those

who expressed a vote intention and the leaning voters, a dummy variable (“Leaners”), coded

1 for leaning voters and 0 otherwise, is included in the regression models.

The dependent variable for the first hypothesis is represented by a binary variable that

registers whether the respondent’s vote choice has changed since the first round. It is scored 1

for vote change and 0 otherwise; 37.9% (N=249) of respondents changed their vote intention

between the two rounds – this includes not only those who switched between the candidates

qualified for the run-off, but also those who voted for other candidates in the first round

and those who did not vote in the first round. For the second, third, and fourth hypotheses,

second-round vote intention (coded 1 for voting Iohannis, 0 for voting Ponta) is the outcome

variable.

For my first, second, and third hypotheses, I use two survey questions5 about exposure

to the first and second debates to construct my main independent variable. Based on these

5 The exact wording of the first question and the possible answers are the following: “The day beforeyesterday evening, Tuesday 11 November 2014, the first presidential debate between Klaus Iohannis andVictor Ponta took place. The debate was broadcast on television, but also on the Internet. Did you watchthe debate? Yes, entirely; Yes, partially; No; Do not know/Do not answer”. The second question has thesame wording, but refers to the debate organised on Wednesday, 12 November 2014. When I collapsed thecategories and built the new debate exposure variable for each debate, those who watched the debate entirelyor partly were grouped in the “yes” category, those who did not watch are placed in the “no” category, whilethe respondents who did not know or did not answer were excluded from the analysis. Note that there is noway to distinguish between those who did not know and those who did not answer.

13

Page 14: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

questions, I create a variable that counts the number of debates each respondent was exposed

to6. Overall, 22.4% watched none of the debates, 21.9% only one and 55.7% both.

For the fourth hypothesis, the main independent variable is the perception of who won

or lost the first and second presidential debates, respectively. For each debate, respondents

were asked: “Which of the two candidates seemed better, more convincing, more pertinent

in this debate: Iohannis or Ponta?”7. I created two variables which register, for each re-

spondent, how many debates were won by Ponta and how many by Iohannis8. Based on

this operationalisation, 62.5% of the sample included in the analysis (N=323)9 believed that

Iohannis won no debates, 15.5% only one debate and 22% both, while 35.5% considered that

Ponta won no debates, 11.4% only one and 53% both10.

As control variables, I employ age (continuous variable), sex (coded 1 for female, 0 for

male), and education (a categorical variable with three levels – primary education or less,

secondary education, and post-secondary education, with primary education serving as the

reference category; those who did not know or answer were excluded)11. These control

6 An alternative measurement option was to have two debate exposure dummy variables, correspondingto each debate. As a robustness check, I performed the analyses with these alternative operationalisations;the results were similar to those reported in the text, whether both dummies were included or only thedummy for the second debate. The results are available from the author.

7 For both debates, the exact wording of the question and the possible answers are the following: “Whichof the two candidates seemed better, more convincing, more pertinent in this debate: Iohannis or Ponta?Iohannis was much better than Ponta; Iohannis was slightly better than Ponta; Ponta was slightly betterthan Iohannis; Ponta was much better than Iohannis; both were as good in this debate; both were as weakin this debate; I did not watch the debate; Do not know/Do not answer.” In accordance with the standardpractice, the respondents who did not know or did not answer were excluded from the analysis.

8 A candidate is considered the winner of the debate if the respondent says that the contender was betteror slightly better than his opponent; otherwise, the variable for that candidate is coded 0. Those who did notwatch the debates are excluded from the analysis. As one can see from the wording of the survey question,only those who were directly exposed to the debate were asked to provide an answer; this means that thequestion captures only the direct effects of the debates (which is a limitation of the analysis) and representsa conservative assessment of the debates’ effect.

9 The explanation for this reduced sample size comes from the fact that the non-watchers were not askedwho won each debate; in addition, I dropped those who did not provide an answer about their educationlevel and their vote in the first round. In Section 5.2, I employ an alternative operationalisation of the mainindependent variables that increases the sample size and confirms the first results.

10 For the fourth hypothesis, I also tested an alternative operationalisation of the main independentvariable. I created one variable for each debate, which was scored 1 if Iohannis was better, 0 – if both wereas strong/weak or if the respondent did not know or did not answer, and -1 if Ponta was considered thewinner. This alternative operationalisation yields similar results (results available on request).

11 Another control variable that I considered was income (a categorical variable with three levels: lowincome, medium income, and high income). However, 49.6% of those expressing a vote intention did not

14

Page 15: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

variables have been highlighted in previous research on Romanian elections (Fesnic, 2010;

2012). Reported vote in the first-round is also included to guard against possible omitted

variable bias since a predisposition in favour of one or other candidate in the second round

could bias perceptions of debate performance. This is in line with the Michigan School

tradition (Campbell et al., 1960), which assumes that party identification has a major direct

and indirect impact on the vote. Reported vote in the first round is a categorical variable

with four levels: those who voted for Ponta, those who voted for Iohannis, those who voted

for other candidates12 and those who did not vote in the first round; the first-round non-

voters serve as the reference category and the respondents who did not know or answer are

excluded. Descriptive statistics for the dependent, main independent and control variables

are presented in Appendix B.

For the first hypothesis regarding debate exposure to be validated, the debate exposure

variable has to have a positive and significant effect on the probability of vote change (i.e.

a difference between the reported vote in the first round and the vote intention expressed

immediately after the two presidential debates). To get at the impact of debate exposure

on vote switching between the two main contenders, I also estimate a model restricted to

those who chose one of the two final candidates in the first round. To assess the second

hypothesis regarding the impact of debate exposure on helping respondents make up their

minds, I estimate another model for those who voted for another candidate or did not vote in

the first round. Finally, to evaluate the third hypothesis, I test a model where the dependent

variable distinguishes between those who had a vote intention for the second round and those

who did not and were still undecided. Given this specification, the debate exposure variable

provide an answer to this question (the DK/NA category). The share of missing data was simply too bigto employ multiple imputation. To deal with this issue, I also created three dummy variables: low income(1 for those reporting a low income, 0 for those with medium and high incomes and DK/NA answers);medium income (1 for medium income, 0 for those with low and high incomes and DK/NA answers); andmissing income (1 for those with DK/NA answers, 0 for the rest). When these variables were includedin the regression models, their effects did not even approach conventional levels of statistical significance.Accordingly, income can safely be excluded.

12 This coding was informed by the fact that, in the first round, there were 15 candidates running, withthe first two obtaining together more than 70% of the votes and the third gaining the support of only 5.36%of Romanians.

15

Page 16: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

would indicate whether the debates helped respondents to make up their minds.

For the fourth hypothesis, regarding the impact of the perceived debate winner on vote

intentions, the expectation is that respondents who consider one of the candidates to have

won the debate are more likely to vote for that candidate. Respondents who voted for other

candidates or did not vote at all in the first round are theoretically the most likely to be

influenced by the debates.

Given that I have binary dependent variables, I test my hypotheses using logistic re-

gression. Since the coefficients of logistic models are not directly interpretable, I estimate

predicted probabilities, using the observed values method (Hanmer & Kalkan, 2013). This

makes it possible to discuss the behavioural effects in a meaningful way.

5 Findings and discussion

5.1 The impact of debate exposure

The regression model displayed in Table 2 allows us to test the first hypothesis. Exposure

to the two presidential debates has a statistically significant effect on vote change (p-value

of 0.06), but the effect is negative. In other words, people who did not watch the debates

were more likely to switch votes than those who viewed them. The predicted probability

of vote change for non-watchers is 37%, compared with 32% for those who watched one of

the debates and 28% for those who watched both. While we cannot rule out the possibility

that committed voters were more likely to watch the debate given the research design we

are employing, a possible interpretation of the negative effect of debate watching is that the

presidential debates had a reinforcement effect, making those who watched them even more

likely to vote for their first round choice.

The reinforcement thesis was directly tested in the Sociopol survey. The survey asked

the respondents about the overall impact of the two debates on their vote behaviour13.

13 The exact wording of the question and the possible answers are the following: “Given the two TV

16

Page 17: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 2: The impact of exposure to the 2014 Romanian presidential debates on vote change

Vote Change Model

(Intercept) −0.24 (0.43)Woman −0.17 (0.19)Age −0.01 (0.01)∗

Education Secondary education 0.11 (0.29)Education Post-secondary education 0.57 (0.32)·

Leaners 1.01 (0.29)∗∗∗

Debate Exposure −0.23 (0.12)·

AIC 681.49BIC 711.88Log Likelihood −333.74Deviance 667.49Num. obs. 568

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, ·p < 0.1

60.4% of the respondents reported that they experienced reinforcement of their previous vote

choice following the exposure to debates, while 39.6% reported no reinforcement. As can be

seen in Table 3, there is a statistically significant relationship between debate exposure and

self-reported reinforcement: only 23% of non-watchers experienced reinforcement (76.9%),

compared with 62% of respondents who watched only one debate and 75% of those exposed to

both debates. These results lend weight to the inference that the main role of the debates was

to reinforce existing vote choices, rather than provoking shifts and realignments. However,

this conclusion needs to be treated with caution, given people’s limited ability to report

reliably on their own motivations.

To further explore the vote change hypothesis and provide a more fine-grained analysis, I

estimated additional models for subsets of respondents. The first model displayed in Table 4

debates and the discussions that you heard on TV or had with your friends and family, what effect has thishad on your voting intention? Convinced me even more to vote for Iohannis, for whom I wanted to votepre-debate; Convinced me even more to vote for Ponta, for whom I wanted to vote pre-debate; Before thedebate I wanted to vote for Iohannis, but the debate convinced me to vote for Ponta; Before the debate Iwanted to vote for Ponta, but the debate convinced me to vote for Iohannis; It did not reinforce my existingconvictions, nor made me reconsider my choice; I believe it convinced me not to go to vote; I believe itconvinced me to go to vote. Before the debate I did not want to go to vote, but the debate convinced me togo to vote; Did not see the debate or any public comments about it; Do not know/Do not answer”.

17

Page 18: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 3: Voting behaviour reinforcement based on exposure to debates

No debates One debate Two debates TotalNo reinforcement 113 53 92 258

76.9% 37.6% 25.1%Reinforcement 34 88 274 396

23.1% 62.4% 74.9%Total 147 141 366 654

22.5% 21.6% 56%Chi2 = 117.8 d.f. = 2 p < .001

estimates the impact of exposure to the debates on the likelihood of voting for the right-wing

candidate in the second round. The sample is restricted to Ponta and Iohannis first-round

voters. The model includes reported vote in the first round, making it possible to look at

the impact of debate exposure on vote switching between Ponta and Iohannis. The results

show that, as expected, having voted for Iohannis in the first round (compared to having

voted for Ponta, which is the reference category) significantly increases the chance of voting

for Iohannis in the run-off. The coefficient for debate exposure is non-significant. In other

words, debate exposure had no effect on the probability of vote switching. Once again, this

points to reinforcement.

In order to see whether the effects of exposure differed depending on first-round reported

vote, I added an interaction term between the debate exposure variable and voting for

Iohannis in the first round to the model (see model 2 in Table 4). The positive sign of the

interaction term suggests that first-round Iohannis voters who watched the debates were

more likely to vote for Iohannis in the second round than non-watchers, but the interaction

term falls well short of conventional levels of statistical significance. Given the complexities

of interpreting interaction models (Brambor et al., 2005), I estimated marginal effects for

the interaction and its constituent terms. The results confirmed that exposure to debates

did not significantly influence vote switching between the two contenders.

The next model tests the second hypothesis. It considers only first-round thwarted voters

and non-voters. For this subsample, I am looking at the impact of exposure to the debates

18

Page 19: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 4: The impact of exposure to the 2014 Romanian presidential debates on the likelihoodof voting for Iohannis in the second round (Ponta and Iohannis first-round voters only inmodel 1 and 2; first-round thwarted voters and first-round non-voters in model 3)

M1 - Pontaand Iohannis voters

M2 - Pontaand Iohannis voters

M3 - Thwarted votersand non-voters

(Intercept) −0.13 (1.55) 0.25 (1.51) 1.07 (0.80)Woman −2.28 (1.14)∗ −2.25 (1.14)∗ −0.52 (0.34)Age −0.04 (0.02)· −0.04 (0.02)· −0.02 (0.01)∗

Education Secondary education −1.36 (1.11) −1.25 (1.11) 0.23 (0.54)Education Post-secondary education −1.42 (1.30) −1.36 (1.31) 0.58 (0.63)Leaners −1.52 (1.19) −1.44 (1.19) 0.33 (0.47)Vote R1 Iohannis 9.37 (1.29)∗∗∗ 8.15 (1.66)∗∗∗

Debate Exposure −0.18 (0.50) −0.70 (0.71) 0.33 (0.22)Vote R1 Iohannis : Debate Exposure 0.95 (0.95)Vote R1 Other Candidate −0.00 (0.35)

AIC 75.70 76.72 223.41BIC 107.68 112.69 248.31Log Likelihood −29.85 −29.36 −103.70Deviance 59.70 58.72 207.41Num. obs. 402 402 166

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, ·p < 0.1

on the likelihood of voting for the right-wing candidate. As model 3 in Table 4 highlights, the

debate exposure variable has a positive effect on the probability of voting for Iohannis. The

effect of the debate exposure variable approaches conventional levels of statistical significance

(p-value =.12). Given the reduced sample size (N=166), this is a sign of the potential

role these high-audience media events played in helping first-round thwarted voters and

non-voters make a voting decision. This can be considered evidence for the relevance of

the televised debates for the voting behaviour of these particular categories of voters. The

predicted probability of voting for Iohannis was 65% for first-round thwarted voters and first-

round non-voters who watched both debates, compared with only 50% for non-watchers. For

first-round thwarted voters and first-round non-voters, exposure to the debates mattered, as

it increased their probability of voting for the right-wing candidate.

Finally, I estimated a model where the dependent variable indicated whether or not the

respondent had a vote intention for the second round; the variable is coded 1 if the respondent

19

Page 20: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 5: The impact of exposure to the 2014 Romanian presidential debates on having avote intention

Decided Voters Model

(Intercept) 1.47 (0.67)∗

Woman −0.45 (0.29)Age −0.01 (0.01)Education Secondary education −0.43 (0.44)Education Post-secondary education −0.56 (0.52)Vote R1 Ponta 2.55 (0.44)∗∗∗

Vote R1 Iohannis 3.22 (0.62)∗∗∗

Vote R1 Other Candidate 0.53 (0.34)Debate Exposure 0.43 (0.18)∗

AIC 351.73BIC 391.81Log Likelihood −166.87Deviance 333.73Num. obs. 635

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, ·p < 0.1

is decided and 0 otherwise. This tests the third hypothesis. In this case, the goal is to test if

exposure to the presidential debates helped voters to decide their second-round vote. As we

can see in Table 5, watching the debates seems to have made respondents more likely to have

settled on their vote choice for the second round, though we cannot rule out the possibility

with cross-sectional data that committed voters were more likely to watch the debates. The

predicted probability of being decided is 84% for non-watchers, compared with 92% for those

who watched both debates. Clearly, exposure to the debates mattered in terms of deciding

whom to vote for in the second round – a validation of the third hypothesis.

At this point, the various analyses have shown that the first hypothesis can only be

partially accepted, while the second and third ones are fully supported. Exposure to the

debates seemed to help people make up their mind in relation to the second-round vote. For

first-round thwarted voters and first-round non-voters, exposure to debates mattered, as it

increased their probability of voting for the right-wing candidate. The debates also helped

respondents form a vote intention for the run-off. However, the debates mattered for those

20

Page 21: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

who voted for Ponta or Iohannis in the first round only by reinforcing their previous vote

choice; exposure to the debates did not significantly influence vote switching between the

two contenders. Despite the limitations of the cross-sectional research design, the findings

suggest a nuanced understanding of the behavioural effects of presidential debates and the

importance of analysing subsets of voters.

5.2 The impact of the perceived winner

The fourth hypothesis is tested by examining the impact of the perceived debate winner

on the likelihood of voting for the right-wing candidate (see Table 6). Among those having

a vote intention (N=657), 44% indicated Ponta as the winner of the first debate (compared

to 19% who preferred Iohannis); Ponta was also considered the winner of the second debate

(33.2%), but the margin over his right-wing opponent (21.3%) was significantly smaller14.

The first model in Table 6 excludes non-watchers because these respondents were not asked

who won each debate. Consequently, the total sample size is only 323. Despite the small

sample size, the effects of the perceived winner are statistically significant: perceiving Ponta

to be the winner of the debates decreases the likelihood of expressing a vote intention for

Iohannis whereas considering Iohannis the winner debates increases the probability of voting

for him. The results hold even when controlling for reported vote in the first round. Someone

who identified no winners has a 50% predicted probability of voting for Iohannis, while a

respondent who perceived Iohannis to be the winner of both debates has a 93% predicted

probability of voting for the right-wing candidate. In the case of those who considered Ponta

the winner in both media events, the predicted probability of voting for Iohannis is only 24%.

This provides strong support for the fourth hypothesis.

As a robustness check, I employed a more conservative operationalisation of the debate

winner perception variables. I coded the respondents who did not watch one or both debates

as persons who said that no candidate won the debates. Although the sample size increases

14 However, it should be kept in mind that the questions capture only the direct effects of the debates –those who did not watch the media events were not asked who won.

21

Page 22: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 6: The impact of the debate winner perception on the likelihood of voting for Iohannisin the second round

Winner perception modelWinner perception model

(conservative sample)

(Intercept) 5.62 (2.56)∗ 1.87 (0.92)∗

Woman −1.51 (1.08) −0.82 (0.43)·

Age −0.03 (0.03) −0.03 (0.01)∗

Education Secondary education −2.68 (1.68) −0.09 (0.59)Education Post-secondary education −1.50 (1.56) 0.63 (0.68)Leaners −1.33 (1.52) 0.07 (0.54)Vote R1 Iohannis 4.73 (1.47)∗∗ 3.59 (0.66)∗∗∗

Vote R1 Other Candidate 0.60 (1.01) −0.01 (0.47)Vote R1 Ponta −5.25 (2.10)∗ −4.01 (0.85)∗∗∗

Ponta Winner −2.14 (0.66)∗∗ −1.34 (0.31)∗∗∗

Iohannis Winner 3.91 (1.30)∗∗ 4.44 (0.98)∗∗∗

AIC 62.87 180.88BIC 104.43 228.64Log Likelihood −20.44 −79.44Deviance 40.87 158.88Num. obs. 323 568

∗∗∗p < 0.001, ∗∗p < 0.01, ∗p < 0.05, ·p < 0.1

to 568, this option is more conservative because it reduces the variance in my debate winner

variables. As can be observed in model 2, there are no major and substantive differences

compared to model 1 – the same variables are statistically significant and they maintain

their sign; moreover, due to the increased sample size, the level of statistical confidence for

the abovementioned variables increases.

At this point, we can conclude that the fourth hypothesis is confirmed. Indeed, voters

are more likely to say they will vote for the candidate who is perceived to have won the

presidential debates. The relationship is strong, holds under two different operationalisations

of the main independent variables and even when predispositions are taken into account, as

indicated by reported vote in the first round.

22

Page 23: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

6 Conclusions and limits of the analysis

This paper represents, to my knowledge, the first investigation of the behavioural effects

of presidential debates in a Central and Eastern European country. In relation to the 2014

Romanian presidential elections, it set out to address four main questions: first, whether

voters who were exposed to the two presidential debates which took place between the two

rounds of the election were more likely to change their vote intentions; second, whether ex-

posure to the presidential debates had a significant impact on the vote intentions of thwarted

voters and first-round non-voters; third, whether exposure to debates increased the proba-

bility that a voter will form a second-round vote intention; and, fourth, whether voters were

more likely to say they would vote for the candidate whom they perceived to have won the

presidential debates.

The analyses provide only partial support for the first hypothesis. The debates had little

impact on those who had voted in the first round for the candidates qualified for the run-

off; for this subset of voters, the role of the debates seems to have been one of reinforcing

previous behaviour. However, the second hypothesis is confirmed: exposure to the debates

mattered for first-round thwarted voters and first-round non-voters. The same can be said

about the third hypothesis: for those who had a decision to make regarding their vote in the

second round, the debates were important. Specifically, someone who watched the debates

was more likely to form a second-round vote intention. In a close election, the votes of these

categories could prove decisive. As Chaffee (1978) argues, debates are most likely to have

an effect on vote choice when the contest appears to be a close one. The impact of the

debates on these voters suggests a new mechanism through which electoral persuasion and

mobilisation can take place. A two-ballot electoral system expands the realm of possibilities

for campaign effects. When looking at the effects of presidential debates in such systems,

researchers need to consider the whole spectrum of less committed voters. As this paper has

shown, there are three broad types of voters who cannot be ignored if the goal is to offer a

comprehensive understanding of the impact of debates in different democratic settings.

23

Page 24: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

I found strong support for the fourth hypothesis. Perceiving a candidate to be the

winner of the debates increased the likelihood of voting for him. The finding holds even

when controlling for the reported vote in the first round and using a more conservative

operationalisation of the dependent variable.

The literature on the behavioural effects of debates is based mainly on research conducted

in the United States (presidential elections) and other Western democracies (federal elections

in Canada and Germany). Less interest has been shown in debates taking place in elections

using the two-ballot system. As the paper has shown, different dynamics operating in the

case of two-round elections point to the need to expand the circle of research. In particular,

the conclusions regarding the first-round thwarted voters and first-round non-voters have

both theoretical and practical consequences; they allow us to move beyond the question

of whether debates matter to the questions about for whom they matter and under what

conditions. For example, in the United States, the debates are more likely to have an

impact on those whose party allegiances are weak or non-existent (Chaffee, 1978) – leaning

Democrats, leaning Republicans, and independents. The existence of two-round elections

in countries characterised by party systems with more than two parties calls attention to

the role played by first-round thwarted voters, first-round non-voters, and, in general, by

those who have to make a decision about their second round vote. These types of electors,

which obviously do not exist in the context of the US presidential elections, are relevant

when assessing debate effects in Central and Eastern European countries. Thus, generalising

the conclusions of studies of the effects of debates drawn from the American, Canadian or

German experiences ignores important and potentially decisive swaths of voters who are a

reality in the electoral processes taking place in new democracies. In other words, existing

studies may have underestimated the overall impact of televised debates between political

leaders because their findings are context bound.

The paper is the first attempt in the region to address this topic. Powerful similarities

support the idea that the findings would apply to other Central and Eastern European

24

Page 25: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

contexts (especially Poland, the Czech Republic, Slovakia, Croatia, and Bulgaria): shared

communist legacies and institutional trajectories after the end of the Cold War (related to EU

and NATO membership); the role of presidential elections as first-order elections; presidents

directly elected in two-ballot systems; the run-off organised two weeks after the first round

of the elections; presidential debates held only a few days before the second-round vote. The

potential for generalisability is clear, but its extent is ultimately an empirical question. Our

conclusions will have to be tested in different CEE countries and in different election years.

Moreover, better research designs (i.e. panel studies) are required to be able to confidently

address the issue of causality, something this paper is unable to do, as it is based on only one

post-debate survey. In particular, given the research design we are employing, we cannot

rule out the possibility that committed voters were more likely to watch the debates. Also,

we need to be sceptical of people’s ability to report reliably on their own motivations and so

the reinforcement thesis has to be treated with caution. More research will shed more light

on these promising findings.

25

Page 26: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

7 References

Blais, Andre; Boyer, Martin M. (1996), Assessing the impact of Televised Debates: The

case of the 1988 Canadian Election, British Journal of Political Science, Vol. 26, No. 2

(April 1996), pp. 143 – 164.

Brambor, T.; Clark, W. R.; Golder, M. (2005), Understanding interaction models: Im-

proving empirical analyses, Political analysis, 14(1), pp. 63 – 82.

Campbell, Angus; Converse, Philip E.; Miller, Warren E.; Stokes, Donald E. (1960), The

American Voter, Chicago: The University of Chicago Press.

Chaffee, Steven H. (1978), Presidential debates – are they helpful to voters?, Communi-

cation Monographs, 45(4), pp. 330 – 346.

Erickson, Robert S; Wlezien, Christopher (2012), The timeline of presidential elections:

how campaigns do (and do not) matter, Chicago: The University of Chicago Press.

Florin Fesnic (2010), Determinanti ai euroscepticismului in Romania [Determinants of

Euroscepticism in Romania] in Comsa, M.; Gheorghita, A.; Tufis, C., Alegerile pentru Par-

lamentul European, Romania 2009 [The 2009 European elections in Romania], Iasi: Polirom.

Fesnic, Florin (2012), Atribute socioeconomice si demografice si impactul lor asupra sim-

patiei pentru candidatii la alegerile prezidentiale din 2009 [Socioeconomic and demographic

characteristics and their impact on the symphaty for the 2009 presidential candidates] in

Comsa, M.; Gheorghita, A.; Tufis, C.D., Alegerile prezidentiale din Romania, 2009 [The

2009 Romanian presidential elections], Cluj-Napoca: Presa Universitara Clujeana.

Gerstle, Jacques (2008), La communication politique [The political communication],

Paris: Armand Colin (2nd Edition).

Gherghina, Sergiu (2015), The Romanian presidential election, November 2014, Electoral

Studies, Vol. 38 (June 2015), pp. 109 – 114.

Grzymala-Busse, A. M. (2002), Redeeming the communist past: The regeneration of

communist parties in East Central Europe, Cambridge: Cambridge University Press.

Hall - Jamieson, Kathleen; Birdsell, David S. (1988), Presidential Debates – The challenge

26

Page 27: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

of creating an informed electorate, Oxford: Oxford University Press.

Hanmer, M. J.; Ozan Kalkan, K. (2013), Behind the curve: Clarifying the best approach

to calculating predicted probabilities and marginal effects from limited dependent variable

models, American Journal of Political Science, 57(1), pp. 263 – 277.

Holbrook, Thomas M. (1996), Do campaigns matter?, Thousand Oaks: SAGE Publica-

tions.

Lanoue, David J. (1991), Debates that mattered: voters’ reaction to the 1984 Canadian

leadership debates, Canadian Journal of Political Science, Vol. 24, No. 1 (March 1991), pp.

51 – 65.

Lazarsfeld, P. F.; Berelson, B.; Gaudet, H. (1948), The people’s choice: how the voter

makes up his mind in a presidential campaign.

Mainwaring, S.; Torcal, M. (2006), Party system institutionalization and party system

theory after the third wave of democratization. Handbook of party politics, 11(6), pp. 204-

227.

McKinney, Mitchell S.; Carlin, Diana B. (2004), Political campaign debates, in Lee Kaid,

Lynda, Handbook of Political Communication Research, London: Lawrence Erlbaum Asso-

ciates Publishers.

McKinney, Mitchell S. (2008), Debates, in Lee Kaid, Lynda; Holtz-Bacha, Christina,

Encyclopaedia of Political Communication, London: Sage Publications.

McKinney, Mitchell S.; Warner, Benjamin R. (2013), Do presidential debates matter?

Examining a decade of campaign debate effects, Argumentation and advocacy, Vol. 49

(Spring 2013), pp. 238 – 258.

Prior, Markus (2012), Who watches presidential debates? Measurement problems in

campaign effects research, Public Opinion Quarterly, Vol. 76, No. 2 (June 2012), pp. 350 –

363.

Schroeder, Alan (2008), Presidential Debates: Fifty years of high-risk TV, New York:

Columbia University Press.

27

Page 28: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Schrott, Peter R. (1990), Electoral consequences of winning” televised campaign debates,

The Public Opinion Quarterly, Vol. 54, No. 4 (Winter, 1990), pp. 567 – 585.

Schrott, Peter R.; Lanoue, David J. (2013), The power and limitations of televised pres-

idential debates: Assessing the real impact of candidate performance on public opinion and

vote choice, Election Studies, Vol. 32, pp. 684 – 692.

Wlezien, Christopher (2010), Elections campaigns, in LeDuc, Lawrence; Niemi, Richard

G.; Norris, Pippa, Comparing democracies - Elections and voting in the 21st century, London:

Sage Publications.

28

Page 29: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

8 Appendix A

For the presentation of the vote intentions and for the empirical analysis, four national

representative surveys conducted by Sociopol, a Romanian polling institute, were considered.

The first survey was conducted between 5-7 November 2014; universe of research com-

posed of individuals aged 18 years and over and made up of 924 respondents; multistage

cluster sampling, with probabilistic selection of the respondents through the random route

technique; method of data collection – Computer-assisted telephone interviewing (CATI),

based on a standardized questionnaire; weighting by sex, age, ethnicity, region, city size;

margin of error of 3.22% at a 95% confidence level.

The second survey was conducted between 8-10 November 2014; universe of research com-

posed of individuals aged 18 years and over and made up of 1,019 respondents; multistage

cluster sampling, with probabilistic selection of the respondents through the random route

technique; method of data collection – CATI, based on a standardized questionnaire; weight-

ing by sex, age, ethnicity, region, city size; margin of error of 3.07% at a 95% confidence

level.

The third survey was conducted between 11-12 November 2014; universe of research

composed of individuals aged 18 years and over and made up of 875 respondents; multistage

cluster sampling, with probabilistic selection of the respondents through the random route

technique; method of data collection – CATI, based on a standardized questionnaire; weight-

ing by sex, age, ethnicity, region, city size; margin of error of 3.31% at a 95% confidence

level.

The fourth survey was conducted between 13-14 November 2014; universe of research

composed of individuals aged 18 years and over and made up of 833 respondents; multi-

stage cluster sampling, with probabilistic selection of the respondents through the random

route technique; method of data collection – CATI, based on a standardized questionnaire;

weighting by sex, age, ethnicity, region, city size; margin of error of 3.4% at a 95% confidence

level.

29

Page 30: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

All four samples were validated based on the 2012 Statistical Yearbook provided by the

Romanian National Institute of Statistics and the 2011 Census.

30

Page 31: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

9 Appendix B - Descriptive statistics

Table 7: Descriptive statistics for the dependent variable

Variable Levels n %Vote R2 (whole sample) Victor Ponta 299 35.9

Klaus Iohannis 274 32.9I would not go to vote 30 3.6DK NA 230 27.6All 833 100.0

Leaning voters Victor Ponta 41 15.8Klaus Iohannis 43 16.5DK NA 176 67.7All 260 100.0

Vote R2 (voters sample) Victor Ponta 340 51.8Klaus Iohannis 317 48.2All 657 100.0

31

Page 32: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 8: Descriptive statistics for the main independent variables

Variable Levels n %Debate1 Exposure No 185 28

Yes 469 71.4DK NA 3 0.6All 657 100.0

Debate2 Exposure No 250 38.1Yes 405 61.6DK NA 2 0.3All 657 100.0

Debate1 Winner Iohannis better 125 19.0Ponta better 289 44.0Both as good/weak 46 7.0Did not watch 188 28.6DK NA 9 1.4All 657 100.0

Debate2 Winner Iohannis better 140 21.3Ponta better 218 33.2Both as good/weak 37 5.6Did not watch 252 38.4DK NA 10 1.5All 657 100.0

32

Page 33: Do presidential debates matter? Evidence from a Central and … · 2017-08-31 · Keywords: Impact of debates, Central and Eastern Europe, Romania, Presidential elections, Voting

Table 9: Descriptive statistics for control variables

Variable Levels n %Sex Male 309 47.0

Female 348 53.0All 657 100.0

Education Primary education 111 16.9Secondary education 358 54.5Post-secondary education 170 25.9DK NA 18 2.7All 657 100.0

Income Low income 196 29.8Medium income 74 11.3High income 61 9.3DK NA 326 49.6All 657 100.0

Vote R1 Victor Ponta 235 35.8Klaus Iohannis 180 27.4Other Candidate 83 12.6Did Not Vote 83 12.6DK NA 76 11.6All 657 100.0

Statistic N Mean St. Dev. Min Max

Age 657 50.422 19.023 18 99

33