Begin Again: Election Campaign and Own Opinions Among First-Time Voters in Romania

Objective. This article analyzes under what circumstances ﬁrst-time voters are likely to be inﬂuenced by the election campaign or by their opinion prior to campaign in their vote choice. Methods. It uses individual-level data from an original online survey conducted among Romanian ﬁrst-time voters in the November 2019 presidential election. Results. The results indicate that those who trust politicians, ﬁnd the campaign informative and use social media are inﬂuenced by election campaign. Those with higher political knowledge and interest follow their precampaign attitudes in casting the vote. Conclusion. The study reveals the existence of distinct causal mechanisms for young voters who are inﬂuenced at the polls by the election campaign or by their own attitude prior to campaign. This indicates the necessity to include the existence of an opinion prior to campaign in future analytical frameworks about voting behavior.

paid attention to as voters. The lack of experience with campaigns can have a dual effect: it can either make first-time voters use the campaign cues to assess candidates (Abrajano, 2005) or decide autonomously on how they vote without being aware of the information benefits provided by the campaign. Second, first-time voters have limited partisan attachment because they could not engage with party activities (legal age) and partisanship is less acquired within the family in recent times (Vraga et al., 2014). Earlier studies indicate that when developing voting intentions, young people may be influenced by discussions with their parents (Zukin et al., 2006), but equally relevant are discussions with older friends who are politically active (Ekström and Östman, 2013) and who might answer them to their question regarding the vote in general or about the political campaign (Harris, Wyn, and Younes, 2010). Also, young voters lack previous exposure to political messages and may develop independent thinking about politics by following the social media (Pasek et al., 2006) or political campaigns. Under these circumstances, the role of campaigns for first-time voters may be different from the general population, and evidence is required to substantiate this role. We know that sometimes the information received during election campaign influence voting behavior, while in other occasions the precampaign attitudes are of crucial importance. However, we know very little about when these two effects occur in the same group of individuals.
This article seeks to address this gap in the literature and tests the explanatory power of these two variables (election campaign and existing attitudes prior to campaign) for the same group of voters: the Romanian first-time voters in the 2019 presidential elections. It analyzes under what circumstances first-time voters are likely to be influenced by the election campaign or by their opinion prior to campaign in their choice at the polls. In the context in which several studies describe young people as being less interested in conventional political participation (Busse, Hashem-Wangler, and Tholen, 2015;Henn, Weinstein, and Wring, 2002), this article is focused on that part of young people still engaged in the voting process (Ekström and Sveningsson, 2017). We propose two sets of explanations. On the one hand, we hypothesize that trust in politicians, the perception of campaigns as being informative and the use of the social media for information purposes will increase the importance of political campaign for first-time voters' decision at the polls. On the other hand, we expect that higher levels of political knowledge, participation and interest will make first-time voters to follow their precampaign attitudes in deciding how to vote. We control for left-right placement, living with parents and medium of residence.
We use individual-level data from an original online survey conducted among Romanian first-time voters in the November 2019 presidential election. The article focuses on first-time voters because this group has not been previously exposed to election campaigns. They are less likely to make assessments based on previous interactions and succumb to information bias (including echo chambers), party identification, or socialization with candidates (Gherghina and Tap, 2020). We are interested only in those respondents who actually voted and for this reason we use a maximum variation sampling strategy (see the section on research design). The survey was launched immediately after the second round of the presidential election so that voters could easily remember the campaign. Our analysis focuses on the presidential election because these are the more popular elections in the country-compared to legislative elections-and campaigns are candidate-centered, being more dynamic and more watched. Romania is chosen as a case study due to the general low level of partisanship over time according to which people often change their preferences (Gherghina, 2014). Further reasons for the case selection were the existence of a relatively heated campaign and the active role of the Romanian youth in recent political events in the country, being well known that younger generations have shown a desire to experiment with different ways of living and doing politics that emphasize creativity as well as freedom and autonomy of action, while distancing themselves from traditional modes of political behavior (Silva and Castro, 2014). Low level of partisanship over time does not mean disinterest, but means that young voters get their information from other sources, such as trusted adults, entertainment, and social media (Sveningsson, 2015).
This article proceeds as follows. The first two sections review the literature, present the general analytical reasoning, and identify several determinants for which the election campaign and own attitudes prior to campaign can influence first-time voters' behavior. Next, we briefly present the research design of this study with emphasis on the case selection, variables, and methodology. The fourth section includes the analysis and interpretation of results. The conclusion reflects on the implications of this analysis for the broader field of campaign effects on voting behavior and discusses avenues for further research.

When Election Campaign Matters
This section argues and tests the extent to which the high importance of election campaigns in the voting behavior of first-time voters can be influenced by information and the attitudes toward the presenter. More precisely, if the citizens trust the source of the message and consider that campaigns include relevant information, they are likely to follow them in casting a vote. As such, we focus on three potential drivers derived from the literature: trust in politicians, the perception of campaign as being informative, and the use of social media for information purposes. We combine arguments from the strands of literature on both the general public and young voters.
First, we argue that citizens who trust politicians are inclined to be guided by campaigns in their voting decisions because they believe in politicians' good intensions and commitment. The argument unfolds as follows. Trust is rooted in politicians' ability to follow the rules of the game (Barber, 1983;Elinder, Jordahl, and Poutvaara, 2015). The belief that politicians try to keep their campaign promises gives a feeling of responsiveness and accountability, which can lead to a positive assessment among voters (Anderson and Brettschneider, 2003), being also a sign of a strong democracy (Shockley-Zalabak, Morreale, and Stavrositu, 2017). Trust in politicians determine voters to watch candidates' campaigns, to listen to their electoral promises and to know more about what candidates can address the issues they cared about most (Schill and Kirk, 2017). In candidate-centered campaigns, there is a process of personalization of politics in which the personal traits of candidates and their messages can influence the electorate (Corner and Pels, 2003;Petrarca, Giebler, and Weßels, 2020).
Voters who trust politicians identify with the candidate, resonate with their personality, and believe what is said in the campaign (McGregor, 2018). They perceive the interests of those whom they trust as being aligned with theirs and by this, candidates support voters' interests (Warren, 2018). Since the idea of being wrong is synonymous with doubt of one's own judgment, trusting the campaign is just the ultimate proof of trusting the right politician. Identification with a candidate generates trust that structures political beliefs and behavior. The more people trust the politician, they accept more easily their campaign ideas and believe that the politician will fulfill campaign pledges (Duval and Petry, 2018;Festenstein, 2020). Voters who trust politicians are usually more connected with them (Vergeer, 2013) and consider the campaign a reflection of politicians' ideas and intentions (postelection pledge fulfillment). In this case, voters think that politicians will fulfill the commitment they have undertaken (Hawley, 2014) because politicians have the right kind of motivations and incentives to be responsive to voters' interests (Festenstein, 2020).
Second, voters' perceptions of their knowledge competence (or information efficacy) to engage in the political process are an important determinant of political engagement and voting (Kaid, McKinney, and Tedesco, 2007). The election campaign is the period when voters can interact with the candidate by participating in specific events or by following their media posts. Young people are likely to vote if they are highly motivated and information during the campaign is one of the motivational factors (Lodge and Stroh, 1993). If the election campaign is aimed at the voters, they feel important, involved, and that influences their choices. When election campaigns emphasize some issues, they increase the importance of these considerations for voters' decisions (Balmas and Tamir, 2010).
An informative election campaign provides people a persuasive stimulant to think about (Ferrín, Fraile, and García-Albacete, 2019;Riker, 1989) to become more interested and engaged (Banducci and Karp, 2003;Hansen and Pedersen, 2014) and to be able to identify candidates' points of view. Earlier research finds that when a candidate and voter share same issue priorities, and those are presented during the campaign, the voter is more likely to vote for that candidate (Abbe et al., 2003). Through all these elements, complex and informative campaigns can influence voting behavior. A large number of young people decide late in the campaign how to vote (Willocq, 2019). To them, short-term determinants are important (Dassonneville, 2016). Voters with a low level of information and political awareness have a higher probability of accepting a message based on campaign cues (Pannico, 2017;Zaller, 1992). Informative campaigns include rich content, which comes in two forms: objective information about facts and policies and subjective information-in the form of cues-that are intended to persuade voters. First-time voters who feel that campaigns are informative will use them as main source of information (Juelich and Coll, 2020) and rarely use alternative sources. A well-documented campaign helps people argue their belief, especially when personalizing messages are used and this is more likely to influence voting (Lawrence et al., 2016).
Third, related to the previous point, political information achieved through social media is much more likely than traditional news channels to influence first-time voters (Wolfsfeld, Yarchi, and Samuel-Azran, 2016). Being easily accessible, the primary source of information for first-time voters about candidates' election campaign is their Facebook account where they are exposed and their opinions are presented in different contexts (Sides et al., 2020). Candidates know how to influence voters and they intensively use online media such as Facebook or YouTube to promote themselves (Gronbeck, 2009). During the election campaign, politicians adjust their profile with images and messages to appear more media friendly (Enli, 2017) and this has a big impact for voters who inform themselves mostly through Facebook (Lee and Oh, 2012). Facebook allows users to glean a more personal perspective of a political candidate (Ha et al., 2013). The candidates have the opportunity to instantly engage voters in a more direct, personal, and potentially interactive manner (Lee and Oh, 2012). Through personal disclosures on Facebook politicians contribute to the individualization and privatization of their political (Lee et al., 2018), they want to get closer to the potential voters, and present common aspects of their lives, as any person has.
Very often, young voters follow politicians on social media because they seek to avoid filtering, framing, and priming of journalists and media outlets (Enli and Rosenberg, 2018). Previous studies have shown that exposure to political information on social media platforms increases first-time voters' vote choice certainty more than exposure to information from offline or nonsocial online media (Ohme, de Vreese, and Erik, 2018). During the election campaign, voters have a greater emotional intimacy with the politician but also evaluate the quality of messages more positively after viewing the Facebook page (Dunn and Nisbett, 2014). The way people interact with politician when they sent messages and read the responses leads to a more positive candidate evaluation and a bigger influence.
As a result of all these arguments, we expect that first-time voters to be influenced in their vote decisions by election campaign if they H1: Trust politicians in general. H2: Consider the campaign to be informative. H3: Use Facebook extensively for information purposes.

When Precampaign Opinions Prevail
This section argues that first-time voters who are knowledgeable, active, and interested in politics are likely to be influenced in their voting decisions by precampaign opinions. The general idea is that voters with such a profile are independent from the information received during election campaigns and they can reach a decision based on what they acquire from experience and other sources. We argue that three potential determinants can play a major role: the level of political knowledge, political participation, and political interest, which are derived from several competing theories that are discussed next.
First, political knowledge is defined as familiarity with major political issues and accurate knowledge of prominent political figures and events (Neuman, 1986), respectively, facts about a political system that an individual can recall from their memory to interpret and understand political events (Clark, 2016). An individual's political knowledge is acquired through political learning processes, which begin much earlier in the individual biography (Abendschön and Tausendpfund, 2017), which is easily applied to first-time voters. Those individuals with political knowledge can form an independent opinion and competently participate in political decision-making processes, including voting (Abendschön and Tausendpfund, 2017). When the level of political knowledge increases, voters are better able to distinguish competitors on the ballot and choose the candidates that match their interests (Jaeger, Lyons, and Wolak, 2017). Political knowledge-acquired over time-allows people to make choices based on their own opinions prior to election campaign (Lau, Andersen, and Redlawsk, 2008). The campaign is only a snapshot and knowledgeable citizens are likely to vote according to what they know in general and not to what is served to them during campaigns.
Young voters acquire political knowledge during the process of political socialization. They are interested about and discuss political news with family, teachers, and friends (Abendschön, 2013). With more knowledge comes more confidence about their capacity to decide whom to vote for before the election campaign. The young voters who hold more political knowledge are also better able to understand the political information they receive and better prepared to transform their political attitudes into action (Shaker, 2012). Young voters who are exposed to political messages, events, or party members are more interested to test the received information against their internalized political predispositions before accepting them. Being in contact with politicians gives voters the possibility to find important information about a party and a candidate long before the election campaign (Cross and Young, 2008). This frequent exposure and interaction with political mechanisms helps them engage more intensely and subsequently leads to higher vote choice certainty.
Second, we expect politically active citizens' efforts to be more likely to vote according to their long-lasting attitudes without considering much for election campaign. On the one hand, political discussions are a relevant type of activity in which individuals discuss the meaning of political messages with other people and are encouraged to express their own opinions and decide for themselves (Hooghe and Dassonneville, 2013). Parents and close friends are considered to be the main socialization agents (Jennings, 2004) and have a remarkable effect on the formation of political opinions (Campus, 2012). Interpersonal communication mobilizes political behavior (Ekström and Östman, 2015) and increases the ability to form a personal opinion. Involvement in political discussions with family or friends plays an important role in youth political socialization as well (Šerek and Umemura, 2015). Political discussions with peers before the election are associated with young people's enhanced voting intentions (Šerek and Umemura, 2015) based on their own opinions rather than on election campaigns.
On the other hand, the active involvement in party activities or contacting politicians helps voters formulating attitudes and opinions as a result of direct interaction with political actors. Through political participation voters learn about political attitudes and behaviors in their social environment. Direct experiences with actors and topics of the elections may increase citizens' vote choice certainty (Austin and Nelson, 1993;Ognyanova, 2020). Because first-time voters have no previous election and voting experience, active political participation might be a useful experience, increasing their feeling of certainty about their vote choice. Talking to others about politics exposes discussants to information they have not been aware of before (Amsalem and Nir, 2019). The feeling of being in contact with political actors and being present in political discussions might help young voters to form an opinion.
Third, political interest is likely to help voters follow events associated with elections long before the campaign starts. Political interest is one of the voters' attitudes that is often invoked in the literature on voting behavior. It is considered the strongest predictor for voting, before other variables such as political information, efficacy, and identification (Levy, Solomon, and Collet-Gildard, 2016;Verba, Lehman Schlozman, and Brady, 1995). It is also a stronger predictor for voting than a cost-benefit analysis (Blais, 2000). Political interest often goes hand in hand with knowledge to determine activity in traditional forms of political participation such as voting (Norris, 2001). Political interest is driven by curiosity and entails a person's predisposition to reengage with content over time (Zeglovits and Zandonella, 2013). Accordingly, voters with high interest in politics actively seek for information and form opinions outside elections. They can form an opinion about which is the best candidate or party in an election (Blais and St-Vincent, 2011). Following all these arguments, we expect that first-time voters are influenced in their vote decisions by their own opinions prior to election campaign if they have H4: High levels of political knowledge. H5: High levels of political participation.

H6: They have high political interest
In addition to these main effects, we control for three variables: left-right placement, living with parents and medium of residence. These are likely to produce an effect on the voting behavior as illustrated by earlier research. 1 First, the positioning of first-time voters in the political space can make them more open toward the campaign or more conservative toward following their opinions prior to campaign when casting their vote.
Extensive research indicates that the ideological orientation of citizens can predict a broad range of value orientations. Second, living with parents could have an influence due to the different partners for political discussions. Voters living with their parents will have more discussions with their family, while those who live on their own are likely to spend more time with their social network; the latter influences the development of political interest (Dostie-Goulet, 2009). Third, voters living in large localities are more exposed to information and have more opportunities to engage politically compared to voters in small localities. For example, in large localities there are more political activities (e.g., protests, demonstration) but also more possibilities to engage in the activities of political parties. This difference in the level of activity may happen because people are more exposed to outdoor political events, not only to those taking place online, which are accessible from anywhere. Otherwise, if citizens are not exposed to it, the information will have no effect (Lefevere, 2016).

Research Design
This article uses individual-level data from an original online survey conducted among Romanian first-time voters in November-December 2019. The survey was launched immediately after the second round of the presidential elections and closed three weeks later. The survey includes 664 young people who voted in the presidential elections. They were born between 1999 and 2001, being for the first time entitled to vote in national elections. The previous national elections were held in 2016 and those born in 1999 had not attained the minimum age of voting (18 years old). Since we focus on young people who actually voted, we use a maximum variation sample. In the absence of official reliable statistics regarding the profile of young voters, we cannot know the features of the entire population and thus no probability representative sampling can be drawn. Nonprobability sampling is often used to study populations where formal access to complete lists of members is not possible. Maximum variation sampling is a purposive sampling technique used to increase the variation on several key variables. In our case, the survey aimed to maximize the variation in terms of first-time voters' attitudes toward politics and political institutions, political participation (including voting), preferences for decision-making processes, medium of residence, and gender. We are aware that such a sampling strategy confines the findings presented in this article to our respondents, that is, no generalization to a broader population. Nevertheless, we consider the results highly informative and with important implications for the study of effects of election campaigns on first-time voters.
The respondents were neither preselected nor part of a pool of available individuals. We distributed the online survey mostly through messages on Facebook groups or discussion forums, and e-mails sent to organizations or associations. The data set includes only the respondents who completed the survey. The questionnaire was in Romanian and the average time of completion was nine minutes. There is great variation in the respondents' profile across all the variables included in this analysis (Table A1) and for other sociodemographic variables not reported in the analysis (e.g., gender has a distribution that resembles the most recent census at national level).

Variable Measurement
This study has two dependent variables. The first captures the importance of election campaign, on the one hand, and on the other hand, the relevance of own opinion prior to campaign, in deciding how to vote. Respondents were asked "When deciding how to vote in this election, to what extent were you influenced by candidates' election campaign?" and "When deciding how to vote in this election, to what extent were you influenced by your own opinion prior to campaign?" The available answers were recorded on a four-point ordinal scale between "not at all" (coded 1) and "very much" (coded 4).
Trust in politicians (H1) is measured with the question "How much trust do you have in politicians." The available answers are recorded on a 4-point ordinal scale that ranges between not at all (1) and very much (4). The perception of a campaign as being informative (H2) is measured with the question "In your opinion how informative was the campaign conducted for these elections?" The answers are recorded on a five-point ordinal scale between not at all (1) and very much (5). The use of Facebook for information purposes (H3) uses the same ordinal scale as the answer to the question "To what extent do you use Facebook for information purposes?" The knowledge (H4) refers to actual or objective as opposed to subjective knowledge. It is a cumulative index based on five questions about Romanian politics that respondents had to classify as true or false. These questions were about the country's E.U. membership, the length of the country president's term in office, the co-habitation between president and prime minister, the bicameral structure of the Romanian Parliament, and whether the government can resign after a vote of no confidence in parliament. All correct answers were coded 1, while the incorrect were coded 0, which results in a 6-point ordinal scale with values between 0 and 5.
The degree of political participation (H5) is measured on a 4-point ordinal scale. It is a cumulative index of three possible actions: voting in a referendum, protests, and signing petitions (both online and offline). The respondents are first-time voters and their possibilities for participation were limited due to age constraints. However, the country organized two referendums in a short interval (one in October 2018 and one in November 2019) and that is why the question refers explicitly to voting in a referendum. The values of the index range between 0 for those respondents who did not participate at all and 3 for those respondents who were engaged in all three activities. Political interest (H6) is measured on a 5-point ordinal scale based on the following question: "How interested are you in Romanian politics." The possible answers range between "not at all" (1) and "very much" (5).
The first control variable is the self-placement of respondents on an 11-point ordinal scale where 0 stands for left and 10 for right. Living with parents is coded on a 3-point ordinal scale where 0 corresponds to living with parents, 1 living independently in the same locality, and 2 living independently in a different locality. Medium of residence is coded on a scale that ranges between 1 (village) and 4 (large city with over 300,000 inhabitants). For all the variables, the "DK/NA" answers are treated as missing values and are excluded from the analysis. The analysis uses ordinal logistic regression to test the effects of the hypothesized effects (Model 1 in Table A2) also including the controls (Model 2 in Table A2). Before running the regression, we tested for multicollinearity and the results indicate no highly correlated predictors, that is, the highest value is 0.30. This is also reflected by the values of the Variance Inflation Factor (VIF) test for multicollinearity, which are smaller than 1.18.

Analysis and Results
Before the analysis, it is useful to provide an overview about the campaign for presidential election in Romania. The country president is elected directly by citizens through popular vote once every five years, in a two-round system. The country presidents are allowed to a maximum of two terms in office. Since the regime change in 1989, all presidents who ran for a second term have been reelected. In general, the campaign lasts for one month before the election and is characterized by media presence, street campaigning, rallies, and different forms of clientelism. Regarding the latter, earlier studies indicate a relatively high incidence of vote buying with money, goods, or welfare promises (Gherghina and Volintiru, 2017;Mares and Visconti, 2019). The campaign ends on the last Thursday before voting, which usually takes place on a Sunday. In 2019, voting by diaspora members took place for three days between Friday and Sunday to provide more Romanian migrants the possibility to vote. In addition to this extended voting schedule, the diaspora also had the possibility of postal voting. The latter was introduced in 2016, following a large protest during the previous presidential election in 2014 when many Romanian citizens could not vote abroad due to the poor organization during Election Day (Gherghina, 2015). 2 In November 2019, the incumbent President Klaus Iohannis ran for reelection and won a second term in office. In the first round, there were 14 candidates representing almost all the parliamentary parties, some extraparliamentary parties or running as independents. Of these, only four candidates were likely to receive more than 10 percent of the votes. The main opponent was the former prime minister of the country, Viorica Dancila, who was dismissed after losing a vote of confidence in parliament one month prior to elections. The election campaign was intense and had a major online component, which matches the focus of this study on first-time voters. President Iohannis favored the online communication with his electorate and many of his actions were often explained through social media. His campaign in 2014 was also conducted heavily online and he maintained that approach five years later. The large majority of his 2019 campaign took place online and the other candidates followed him. Iohannis refused a TV debate with Dancila after they both made it into the second round, explaining that he did not want to engage in a dialogue with the candidate of a party that governed against Romanians for three years (Digi24, 2019a). This message is rooted in the rough cohabitation period that Iohannis had with Dancila's party after the 2016 legislative elections.
According to official statistics, around 700,000 young voters aged between 18 and 24 went to the polls in both rounds of the 2019 presidential elections (Website Central Electoral Bureau, 2019). This is less than 10 percent of the total turnout in each of the two rounds. Some of these voters went for the first time to the polls and they were included in the survey conducted for this study.

Descriptive and Bivariate Statistics
The distribution of respondents according to the influence of election campaign and own opinion before the campaign (Figure 1) provides some useful insights into the profile of the surveyed first-time voters. The horizontal axis depicts the percentage of respondents, the gray bars reflect where they are positioned on the election campaign, while the black bars depict the extent to which own opinions influenced their vote choice. The election campaign is quite important in the decision to vote for 45 percent of the respondents but plays a small role for 30 percent. The opinion formed before the campaign matters much for 45 percent of the respondents and very much for roughly 35 percent. Overall, a large  share of the respondents is influenced either much or very much by the campaign or their own opinion. In spite of the similarity at aggregate level, the respondents are distributed differently across the two variables. The correlation between the answers provided by respondents to the two questions is 0.11, statistically significant at the 0.01 level. This low value of the coefficient indicates that they are distributed in different categories (the bars in the graph) across the two dependent variables. In other words, the election campaign and own opinion before the campaign have a different effect for most first-time voters included in our survey. The bivariate correlations in Table 1 indicate the existence of empirical support for five out of the six hypothesized relationships. The values of the correlation coefficient indicate that the first-time voters who trust politicians (H1), perceive the campaign as being informative (H2) and use Facebook for information purposes (H3) are more likely to be influenced by the content of election campaigns in their vote decisions. In addition to these hypothesized relationships, the bivariate analysis indicates that first-time voters with high levels of political interest are also more likely to consider the campaign when deciding how to vote. Among the controls, voters placed more to the right are also likely The Effects on the Political Campaign and Own Opinion Prior to Campaign to be influenced by the election campaign. In Romania, the placement to the right does not reflect conservatism but is a direct reaction to the left orientation of the Social Democrats, Dancila's party, which is a prominent presence on the Romanian political scene in the last three decades. Young people usually oppose the social democrats in Romania, illustrated both through real-life examples of protest and voting but also through the average for our respondents on this variable (6.47; see Table A1).
The correlations indicate also that voters with higher political knowledge (H4) and political interest (H6) are more likely to follow their opinion before campaign. Both are statistically significant at the 0.01 level. The level of political participation is positively correlated with own opinion, but weakly and without statistical significance. The positive perception of campaign is also positively related to following own opinion, which means that voters can assess the content of campaign although their preestablished opinions matter a lot in the voting decision. These results suggest that those first-time voters who are influenced by election campaigns in their voting decisions have a relatively different profile than those who follow their own opinions. The former rely extensively on information and trust, while the latter are driven more by knowledge and to some extent participation; their shared feature is political interest, which is not surprising since this usually characterizes active voters to a great extent. Figure 2 depicts the effects of the six independent variables on the two dependent variables. We run separate models that are presented with different colors: gray for election campaign and black for own opinion. We present here only those models without controls, while the full model specifications are available in Table A2. The results of the ordinal logistic regression confirm to a great extent the observations from the bivariate analysis and provide empirical support to five hypothesized relationships. The effects for which we find empirical support are statistically significant at either the 0.01 or the 0.05 levels. The levels of statistical significance are partially influenced by the low number of cases, but that is beyond our control. There are two important general observations when comparing the effects for the two dependent variables. First, the direction and size of effect differs greatly between them. For example, none of the variables that influence election campaign (H1-H3) have either a strong or a statistically significant effect on own opinion. Second, even for a variable with limited effect, for example, political participation (H5), the effect goes in a different direction for the two dependent variables.

Understanding the Effects
The regression model for the first-time voters who are influenced by election campaign indicates that the strongest effect can be observed among those who trust politicians (H1) and who perceive the campaign as being informative (H2). These are on average 1.4 times more likely to be influenced by the election campaign when voting. Young Romanian voters who have a greater trust in politicians also manifest a greater interest in politics and have higher levels of curiosity regarding the political campaign content. For example, several young Romanian voters (high school students) stated in the context of the last presidential elections, that "I feel that my vote counts. To be a competent, involved president and to put the country in a good light," "Young people should vote. Today we are one click away from any source of information and then we have to create our own opinion from a political point of view," or "I felt I had the power to do something in this country and I felt like a real citizen" (ProTv, 2019). After the vote, the results showed that the winner (Iohannis) obtained more votes from the age category where first-time voters are included, although in the first round the most votes from this category were obtained by the candidate ranked on the third place (Digi24, 2019b). Consequently, the election campaigns with a higher informative content including specific references to young people, determines them to vote, thus confirming earlier findings, and to be influenced in their voting decision. This happens because there are more chances for them to identify with the promoted messages and it also increases their confidence in the electoral program.
A weaker effect is observed for H3 where voters who use Facebook for information purposes are almost 1.2 times more likely to be influenced by election campaigns. One explanation for this effect is that much content of the 2019 presidential campaign in Romania was featured on Facebook. Social media users were exposed more to its content than those young voters who got their information from other media where the campaign was less prominent.
The regression model for the first-time voters who are influenced by their own opinion show a strong effect for political knowledge (H4) and interest (H6). On average, first-time voters with high levels for any of these two are roughly 1.3 times more likely to follow their own opinion when deciding how to vote. Political interest is also a strong predictor for election campaign but loses statistical significance when controls are introduced (Model 2). One explanation for which politically knowledgeable and interested first-time voters follow their own opinion of voting is that they acquire information proactively. The vast majority of the surveyed first-time voters have high school education-even if they are in their junior year at university-and the Romanian educational system does not include courses on politics. In order to achieve objective knowledge about politics, these voters should have actively looked for it, shown genuine interest and thus form an opinion. 3 Another explanation is provided by earlier findings. Young people who actively search for political information on the Internet demonstrate a higher interest which might helped them develop a better political knowledge because they develop a particular reason to think about political systems, elections or candidates (Cantijoch, Jorba, and San-Martin, 2008). Good knowledge about politics provides young voters with different ideas and interpretations of topics in which they have an interest, making them more difficult to be influenced by election campaign.
Political participation has a small negative effect on election campaign but none on own opinion. One possible explanation for this result is that many young people were politically active before the 2019 presidential election. For example, in 2018 there were several protests to which young people participated, while in the spring of 2019 there were the European elections and a referendum where young people voted more than usual (Gherghina and Tap, 2020). During the 2019 elections for the European Parliament, the young voters in Romania were the target audience of several online campaigns about the importance of political participation. None of the three control variables has a significant effect on the two dependent variables. The only strong effect is the medium of residence where first-time voters living in villages are slightly more likely to be influenced by election campaign (OR = 0.91), while those in large cities by their own opinion (OR = 1.11).

Conclusions
This article analyzed under what circumstances first-time voters in the 2019 presidential elections were likely to be influenced by the election campaign or by their opinion prior to campaign in their vote choice. This is one of the few studies seeking to observe these effects on the same group of voters. The results indicate important differences in terms of effects. High trust in politicians, the perception of campaign as being informative, and the use of Facebook for information purposes increase the likelihood of a vote according to the content of campaign. These variables have much more limited effect on the use of the own opinion prior to campaign when reaching a decision. The young voters with more political knowledge and with more political interest are likely to be influenced in voting by their own opinion. The degree of political participation has no effect with respect to the extent to which young voters follow their own opinion, but political active young voters will disregard the content of campaigns.
The implications of these results reach beyond the case study analyzed here. At theoretical level, the study reveals the existence of distinct causal mechanisms for young voters who are influenced at the polls by the election campaign or by their own attitude prior to campaign. The two groups are driven by different variables, and the effects overlap to a very small degree. Such a finding indicates the necessity to include the existence of an opinion prior to campaign in future analytical frameworks about voting behavior. This is of particular relevance to young voters who were not exposed to previous election campaigns. Moreover, our findings engage directly with the extensive literature that describes young people as being less interested in conventional political participation. Our findings nuance this perspective and identify two different categories of young voters: those who trust politicians and follow the campaign around the elections and those who are in general knowledgeable and interested in politics. Thus, instead of referring to generally disinterested young voters, our study reveals that there are particular determinants that make the election campaign or own attitudes to prevail.
The empirical implication lies in the identification of factors that determine voters to choose campaign content over their own opinion or the other way around. With the exception of political interest-which favor the presence in both groups-the other determinants are particular and their effect differs in strength and sometimes in direction.
The findings reveal the existence of a novel pattern among the voters, which can be useful also for policymakers in general. For example, political parties will know that they can influence voting behavior with a solid campaign when their candidates are trusted, when the content of campaign is informative and when voters use extensively social media for information purposes. The first and the third feature are usually known prior to campaigns and thus political actors can easily shape their actions accordingly.
This study is confined to the survey respondents, which brings two inherent limitations. First, the sample size was not very large, and this could have an effect on the statistical significance. The sampling strategy was nonrepresentative and thus not many generalizations can be made. Second, we cannot ensure that respondents were objectively truthful in their responses and investigation of subconscious mechanisms of opinion formation would have significantly exceeded the realm of this article. Future research could address this point by conducting qualitative semistructured interviews that can inform how opinions were formed and whether there is a direct relationship to the content of election campaigns. Another avenue for future research could be a comparison between the attitudes of firsttime voters and those displayed by the rest of the population. This will allow observing the particularity of attitudes among those entering the electorate for the first time and will help drawing conclusions about the importance of campaigns for future electoral competitions.