Sailing Uncharted Waters with Old Boats? COVID-19 and the Digitalization and Professionalization of Presidential Campaigns in Portugal

Abstract

This article investigates the extent to which the COVID-19 pandemic fostered significant shifts in election campaigning. The argument is that COVID-19 might have had an impact on both digitalization and professionalization, which might have been regarded as necessary strategies to curb the difficulties brought about by the pandemic. We apply a most similar systems design with a threefold comparative scheme in order to capture and isolate such effects in the campaigns preceding the 2021 Portuguese presidential elections, using data from campaign spending, campaign activities, and social media activity and impact. Results show that the pandemic crisis has not, generally speaking, brought about a higher level of digitalization of electoral campaigns, in spite of online events having become more common. On the contrary, while there were signs of feebler patterns of normalization of online competition in 2021 vis-à-vis 2016, namely in terms of engagement, normalization was stronger after the lockdown than before. Lastly, relative investment in professionalization was similar in 2016 and 2021, and the difference between the budgeted and the actual investment in 2021 cannot be attributed to the worsening of the pandemic situation or to the lockdown. In sum, we depict a scenario of remarkable stability of the electoral campaigns put forward by presidential candidates in terms of digitalization and professionalization. Its possible causes and consequences are discussed.

1. Introduction

Political actors tend to be creatures of habit, but the most successful ones are those who able to quickly adapt to new environments. The rapid spread of COVID-19 infections since late 2019 was a crucial test to the political parties and candidates’ ability to adapt. Amongst other challenges, they had to carry out electoral campaigns in harsh and unprecedented conditions. Indeed, the pandemic not only placed the coronavirus and its impacts on campaign agendas, but also constrained the campaign strategies that could be implemented due to its severity and associated uncertainty (Virtosu 2021), in some cases leading to new ways of campaigning (e.g., Spinelli 2020; Sullivan 2020).

This article focuses the latter phenomenon. Our aim is to contribute to the understanding of how unprecedented contexts, such as the one the COVID-19 pandemic brought about, shape election campaigns. In doing so, we speak to a broader but still underdeveloped line of research (Virtosu 2021; see also Bach et al. 2021; Ferrinho Lopes 2023). Our starting point is the shared assumption that the contexts in which campaigns take place constitute both a threat and an opportunity, leading to adaptations aimed at facilitating the achievement of established goals, and that short-term and/or unexpected events can have a considerable effect on the actions and means used to interact with voters (Norris 2000; Gibson and Römmele 2001, 2009; Schmitt-Beck and Farrell 2002).

Our argument is that the pandemic led candidates to try and overcome the obstacles created by it through a stronger investment in professionalization and digital communication, and that the competition between political actors on social media became more even. To test these expectations, we resort to a particularly broad range of innovative data on campaign spending (budgets and expenses), legacy media presence (TV and newspapers), on-site and online activities, and social media performance (activity and impact) in the months before the presidential election that took place on 24 January 2021, in Portugal, and the immediately preceding comparable race (2016), thus covering 17 individual presidential campaigns. The focus on the Portuguese case is particularly adequate for an analysis of COVID-19 impacts on campaign modes due to the context in which the 2021 campaigns took place—increasing numbers of infections and deaths within the most worrisome COVID-19 wave in the country at the time (Ritchie et al. 2021). The situation was so severe that a lockdown had to be called three days after the official campaign started.

The comparative framework adopted in this article is threefold. First, we compare the 2016 and 2021 presidential election campaigns taken as a whole in order to spot differences between campaigns taking place outside and within the pandemic context. A series of similarities between the 2021 and the 2016 elections (level of competitiveness, timing, moment in the first-order electoral circle, leading candidate and two other candidates rerunning, etc.) allows for a robust analysis of the effects of the pandemic. Second, we carry out a comparison between the three candidates who ran in both elections, in order to rule out the possibility of differences between 2016 and 2021 being due to different pools of candidates. Third, we resort to the lockdown called on 13 January 2021, to compare the very same candidates before and after that event. This “lockdown effect” test is strengthened by a comparison of similar moments in the 2016 campaign, to rule out the possibility of pre/post lockdown differences being actually due to different modes of campaigning in specific stages of the campaign cycle. By adopting this approach, we offer the most empirically robust test of the impact of the COVID-19 pandemic on campaigns to date.

The article is structured as follows. The next section discusses the literature on elections and election campaigning during the pandemic and presents the expectations derived both from that literature and from extant knowledge on electoral campaigns. After that, we lay out the main features of the cases under study, as well as the variable operationalization strategy. The empirical findings are then presented and discussed. The article ends with an analysis of the implications of our findings.

2. The COVID-19 Pandemic and Election Campaigning: Which Influence on Digitalization and Professionalization?

Campaigns evolve in response to technological and societal developments (Norris 2000; Schmitt-Beck and Farrell 2002), yet short-term factors such as external shocks can play a major role on the way campaigns are implemented (Gibson and Römmele 2001, 2009; Schmitt-Beck and Farrell 2002). To what extent did the COVID-19 pandemic act as such an external shock to parties and candidates, leading them to change the way campaigns are carried out?

To date, there are but a handful of studies that explicitly focus on the nature of electoral campaigns during the pandemic. Looking at the French local elections, Bach et al. (2021) note that the closure of schools, restaurants, and non-essential business activities led to a reduction of in-person campaign actions. Sullivan (2020) also reports a decrease in on-site activities during the 2020 US presidential campaign and a greater importance of online activities, namely those carried out through videoconferencing. In a similar vein, digital campaigning was spotted in the 2020 legislative election campaign in South Korea (Spinelli 2020; Virtosu 2021). Altogether, this evidence points to a trend of growing digitalization—that is, an increase in the importance of online communication (Blumler 2016; Grusell and Nord 2020)—of electoral campaigns, which can be associated to the pandemic.

Digitalization is an important trend in 21st-century campaigns (Grusell and Nord 2020; Sampugnaro and Montemagno 2021). Since the late 1990s, the features of the so-called modern campaign style—traditional media presence and campaign activities aimed at fostering media attention—have been losing space to digital campaigning, which encompasses innovative forms of interactive and targeted communication between candidates and voters powered by the internet at large and the social media in particular (Norris 2000; Schmitt-Beck and Farrell 2002; Vergeer et al. 2013). In Portugal, however, extant research shows that digitalization has not been a key feature of campaigning, as most electoral campaigns are closer to the modern type, with a central role of traditional media and a modest use of online platforms (Seiceira and Cunha 2015; Magalhães et al. 2020; see also Santana-Pereira 2022). This is a remarkable finding in a country with 10 million citizens and 8.5 million social media users, and where the average person spends almost 2.5 h on social media every day (Data Reportal 2022).

A recent study (Ferrinho Lopes et al. 2023) showed that the 2021 presidential election campaigns were marked by a nonnegligible use of social media and some online events. In spite of that, only three out of the seven individual campaigns were close to the digital campaign ideal-type, with a stronger weight of online events and social media vis-à-vis traditional campaigning. A descriptive report of the 2021 presidential election also suggests that digital campaigning was important in that context, although its intensity varied between candidates (Serra-Silva and Santos 2022). We do not know, however, if the pandemic did indeed play a role by significantly leading to increased digitalization vis-à-vis the pre-pandemic world, as both pieces of research mentioned above are descriptive and focused on a single time-point.

In order to shed light on this, and based on the surveyed literature, we expect that the pandemic context promoted a wider use of online events and social media by the Portuguese presidential candidates in 2021 vis-à-vis what happened in the previous presidential election (H1a). Moreover, when looking at the three candidates running in both elections, we expect a stronger weight of digital campaigning in 2021 when compared to 2016 (H1b). Lastly, as on-site events became further restricted, digitalization should be more pronounced after the January 2021 lockdown was declared (H1c).

Our second set of hypotheses can still be placed in the realm of digital campaigning, as they have to do with the influence of the pandemic context on the patterns of online competition between candidates on social media.1 In this type of analysis, the contrasting concepts of equalization and normalization are often used. The equalization hypothesis posits that digital platforms benefit first and foremost parties and candidates who lack resources or have feeble electoral potential, as they will be able to overcome the disadvantages typically faced in the offline environment (with the legacy media often not willing to devote too much airtime or space to them) and communicate their message to a wider audience more effectively (Gibson and McAllister 2015, p. 529; see also Bimber and Davis 2003). Indeed, research has found signs of equalization first and foremost in countries characterized by restricted legacy media environments matched with an unrestricted online media panorama (Strandberg 2008). In turn, normalization is a pattern of competition wherein the main political actors adopt a more sophisticated online presence, as they dispose of more resources, and therefore are more successful than smaller political actors also in the online sphere (Margolis and Resnick 2000; Gibson and McAllister 2015). The Portuguese case has been characterized by a pattern of normalization, with the digital environment being dominated by the main parties and candidates (Cunha and Lobo 2015; Pina 2018; see Santana-Pereira 2022).

We could expect a weakening of the normalization pattern due to the COVID-19 pandemic. Indeed, those candidates who are less able to be in the traditional media had incentives to make further efforts ensuring an intense presence and engagement on social media, as the pandemic context prevented them from trying to overcome that obstacle with on-site campaign activities. Interestingly, a descriptive account of the 2021 presidential campaigns, in which social media data for the 22 days preceding the election was analyzed, identified a trend of normalization in terms of activity (with candidates with higher budgets being more active in the social media) but not in terms of impact/engagement (Ferrinho Lopes et al. 2023). Again, this study does not shed light on the concrete impact of the pandemic. It can be the case that the relationship between campaign spending and online activity was considerably stronger before the pandemic, or that there was also a relationship between finances—or electoral appeal, not analyzed in that research—and impact that vanished in 2021. Additionally, their surveyed evidence comprised only the 22 days preceding Election Day, thus not shedding light on possible shifts throughout different stages of the campaigns.

Based on the aforementioned arguments, our expectations are that patterns of normalization in terms of activity were slimmer in 2021 when compared to 2016 (H2a) and that in 2016, there was also a pattern of normalization in terms of impact that disappeared in 2021 (H2b). We also hypothesize that normalization trends in terms of activity (H2c) and impact (H2d) were milder after the lockdown was called.

Beyond digitalization, in this article we also look into a second key feature of 21-century campaigns (Grusell and Nord 2020): professionalization. Indeed, campaign professionalization has been a hot topic on campaign research over the last 25 years (e.g., Norris 2000; Gibson and Römmele 2001, 2009). This has led not only to an array of conceptual and operational proposals but also to conceptual stretching (Ostrá 2021; see also Lisi 2013), with features of mediatization and digitalization being sometimes included under the umbrella of professionalization (e.g., Gibson and Römmele 2001, 2009; Tenscher et al. 2012; Tenscher 2013). In fact, professionalization and digitalization are not intrinsically connected, as campaigns can be very professionalized and poor in digital terms or very digitalized but without a meaningful investment in professionalization (Grusell and Nord 2020). In this article, we define professionalized campaigns merely as those in which there is a professional team analyzing the market and advising the candidates on specific policy positions, enabling the creation of effective vehicles of communication, and providing training for those directly involved in communicating the message (Lilleker and Negrine 2002; Strömbäck 2007; Grusell and Nord 2020).

In Portugal, in spite of a growing trend in the last decades, the professionalization of political campaigns is still modest (Lisi 2013; Lisi and Santana-Pereira 2015; cf. Santana-Pereira 2022). In 2021, professionalization was less relevant in the budgets of most presidential candidates than events or communication, and three out of the seven candidates did not even spend money on professionalization (Ferrinho Lopes et al. 2023). Yet, that does not mean that the pandemic had no effect. Indeed, there are reasons to assume that the environment brought about by COVID-19 might have enhanced the investment in campaign professionalization. Those reasons stem from the literature on electoral participation and voting behavior during the pandemic. In what regards participation, researchers documented a general trend towards lower levels of turnout, associated with objective or perceived risks of contagion (James and Alihodzic 2020; Santana et al. 2020; Chirwa et al. 2022; Picchio and Santolini 2022). This literature thus identifies an important threat arising from the pandemic context and likely to foster further investment in professionalization—the possibility of voter demobilization. Concerning voting behavior, there is plenty of evidence of increased electoral support for incumbents resulting from the pandemic situation, particularly in contexts where they actively tried to stop the spread of the virus (e.g., Giommoni and Loumeau 2022; Morisi et al. 2021; Leininger and Schaub 2020; Bol et al. 2021; Ferrinho Lopes 2023). In terms of campaigning, this phenomenon raises the expectation of more professionalized campaigns by non-incumbents (i.e., almost all in the context of a race for a personal office such as the presidency), aiming at dealing with this additional incumbency advantage brought about by the pandemic.

In short, since the pandemic context changed the dynamics of interaction with the electorate and created patterns of demobilization and an incumbency advantage that might be seen as threats by most candidates, we anticipate that presidential candidates in general might have allocated a greater deal of their financial resources to professionalization in 2021. In more concrete terms, we expect that the 2021 presidential campaign will be characterized by a stronger relative investment in professionalization when compared to 2016 (H3a). We hypothesize the same trend when looking at individual candidates running in both elections (H3b).

Ferrinho Lopes et al. (2023) also showed that, in general terms, there was an increase in professionalization spending vis-à-vis what had been initially forecast when the 2021 presidential candidates prepared their campaign budgets (although a fine-tuned analysis revealed that this happened only in the case of three out of the seven candidates). Nevertheless, this is no evidence that the increase in professionalization spending was due to the worsening of the pandemic. Indeed, it is possible that candidates are bad at anticipating the real magnitude of professionalization costs and end up spending comparatively more than they expected—something we can only find out by comparing budgets and expenses regarding the 2016, COVID-free, campaigns. As we expect an impact of the pandemic context, we thus hypothesize that the effective investment on professionalization by the 2021 presidential candidates was higher than previously planned, and that a similar trend was not observed in 2016 (H3c).

3. Methodology

This article reports a comparative analysis of presidential campaigns before and during the COVID-19 pandemic in Portugal. In this country, presidential elections are held to elect the Head of State, as the country adopted a semi-presidential regime after the 1974 Carnations Revolution: executive powers are held by a prime minister appointed after legislative elections. The president is not, however, devoid of political powers (see Neto and Lobo 2009; Freire and Santana-Pereira 2019). This was recently emphasized both by the political crisis following the 2015 legislative elections, in which the president played an important role in finding a solution (Fernandes and Jalali 2017), and the pandemic, since it was the president’s constitutional prerogative to call the State of Emergency necessary to implement lockdowns and other restrictive measures.

In the research reported here, we follow a most similar systems design (Przeworski and Teune 1970), by looking at two very similar electoral scenarios. In effect, both the 2016 and 2021 elections took place in the same country, under the same institutional framework, were aimed at filling in the same political office, and were even held on the same date (Sunday, 24 January). Moreover, a subset of candidates (Marcelo Rebelo de Sousa, Marisa Matias, Vitorino Silva) participated in both races. These were also two remarkably uncompetitive elections, since Marcelo Rebelo de Sousa’s victory was foreseeable on both occasions: in 2016, this was due to the candidate being the most notorious and popular (Fernandes and Jalali 2017); in 2021, he was advantaged by the fact that he was running for re-election (incumbent presidents have always been re-elected for a second term in Portugal). Lastly, these two elections took place more or less at the same point of the first-order electoral cycle: in 2016, circa four months after the legislative election, in full honeymoon period; in 2021, sixteen months after the most recent legislative election, but in a context of extended honeymoon due to a pandemic rally-around-the-flag effect (Silva et al. 2021). In short, these electoral contests are as similar as two real-world elections can be.

The distinctive feature is the pandemic. The 2016 race offers the most recent examples of Portuguese presidential campaigns without the pandemic factor. In turn, the 2021 campaigns took place during the most violent COVID-19 wave in the country until then, with a peak of 15,000 new cases and 300 daily deaths (Ritchie et al. 2021). Despite the number of infections being lower when compared to that observed, for instance, during the parliamentary election campaign held one year later (Ferrinho Lopes 2023), death rates during the 2021 presidential campaign have never been reached again, and were amongst the highest around the world at the time (Carvalho 2022). This led the government to declare on 13 January—three days after the beginning of the official campaign period—a mandatory full lockdown to be enforced from 15 January to 30 January. Candidates, voters and institutions were hardly prepared for this: in early 2020, a working group of public entities was created to study concrete policies to curb the effect of the pandemic in the elections, but no guidelines were produced, leading to a climate of public uncertainty and controversy and even to the suggestion that the election should be postponed (for an account of this debate and how the election was carried out, see Luís 2021 and Serra-Silva and Santos 2022).

In addition to the general 2021 vs. 2016 comparison, we also compare the three candidates running in both elections2 and resort to the 2021 lockdown to compare campaign features before and after this restriction was enforced (using also comparable data from the 2016 pre-campaign and official campaign periods3 to strengthen the robustness of the findings). The data used in this article covers the two months before the presidential election days (i.e., from 22 November to 22 January in both cases, given the coincidently same election date in 2016 and 2021).

The seven presidential candidates running in the 2021 election were, by order of votes received: the incumbent center-right Marcelo Rebelo de Sousa, officially endorsed by PSD (Partido Social Democrata; Social Democratic Party)4 and CDS-PP (CDS-Partido Popular; CDS-People’s Party), who was a favorite in pre-election polls5 and ended up receiving 61 per cent of the popular vote; Ana Gomes, an eminent figure of the governing PS (Partido Socialista; Socialist Party), albeit endorsed by LIVRE (Free) and PAN (Pessoas-Animais-Natureza; People-Animals-Nature) due to the informal support of her party to the incumbent president; André Ventura, leader of the populist radical right Chega (Enough); Marisa Matias, a Member of the European Parliament (MEP) from BE (Bloco de Esquerda; Left Bloc); João Ferreira, an MEP endorsed by the Communists and the Greens (PCP-PEV, Partido Comunista Português and Partido Ecologista ‘Os Verdes’; Portuguese Communist Party and Ecologist Party ‘The Greens’); Tiago Mayan Gonçalves, backed by IL (Iniciativa Liberal; Liberal Initiative); and Vitorino Silva, leader of the microparty RIR (Reagir, Incluir, Reciclar; React, Include, Recycle). Serra-Silva and Santos (2022) offer a detailed depiction of these candidates, their results at the polls and the broader electoral context.

Five years before, along with Marcelo Rebelo de Sousa, Marisa Matias, and Vitorino Silva, seven other candidates participated in the presidential race: three from the Socialist sphere (although the party did not formally support any of them)—António Sampaio da Nóvoa, Maria de Belém Roseira, and Henrique Neto; the Communist/Green candidate Edgar Silva; and three other independent candidates (Paulo de Morais, Cândido Ferreira, Jorge Sequeira) who altogether were not able to gather three per cent of the popular vote.(for more details on these candidates and this presidential election, see Fernandes and Jalali 2017).

3.1. Variable Operationalization

3.1.1. Digitalization

In this article, we use objective measures of digitalization. Concretely speaking, we looked into the activities of the candidates during the two months preceding each presidential election and distinguished between traditional and digital ones. We believe this is advantageous vis-à-vis just using data from candidate interviews or surveys (e.g., Grusell and Nord 2020; Sampugnaro and Montemagno 2021), as the latter tend to offer merely subjective measures of the digitalization phenomenon.

In order to capture the relative magnitude of traditional campaigning, we looked into campaign initiatives that could be associated with the modern campaign model (Norris 2000; Schmitt-Beck and Farrell 2002), namely the presence of candidates in the traditional media and on-site campaign activities (which in Portugal are notoriously aimed at capturing the attention of the media; Santana-Pereira 2022). We are aware that the presence in the legacy media as a whole is not strictu sensu a campaign activity, and we deal with that by not focusing on mere depictions of the candidates by the media but on programs or printed press initiatives the candidates chose—and perhaps pressured—to participate in. Legacy media presence was therefore operationalized through media content analysis, through which we counted, for each candidate in each election, the number of free TV ads (tempos de antena) broadcast by the public channel during the official campaign, the number of participations in televised debates on generalist TV channels (RTP, SIC, and TVI), and the number of interviews given to the same television networks, as well as to eight main Portuguese newspapers and newsmagazines (Público, Jornal de Notícias, Expresso, Sol, Correio da Manhã, Observador, Diário de Notícias, Visão, and Sábado). Altogether, this encompassed the analysis of prime-time television broadcasts for three different channels in 124 days (circa 1488 h of TV) and 653 newspapers and news magazines available online or at Lisbon’s Hemeroteca (Newspaper Library).

In turn, on-site campaign activities were identified via the official information on campaign activities that candidates deliver to the Political Accounts and Financing Entity (Entidade das Contas e Financiamentos Políticos; ECFP)6 until six months after Election Day. In specific, we counted the number of rallies and gatherings, outdoor campaign initiatives (such as in-person canvassing and door-to-door contacts), and visits to specific institutions (schools, factories, hospitals, etc.) in which the presidential candidates engaged in.

Regarding digital campaigning, we focused on online events and social media activity. For online events, we identified and computed the number of online fora, presentations, debates, or conversations broadcast on the candidates’ social media accounts and official websites.7 Social media activity was operationalized via the number of publications candidates made on two mainstream and widely used social media in Portugal (Facebook, with 6.3 million users in 2021, and Twitter, with 1.4 million users; Data Reportal 2021). In the case of Twitter, the information was automatically extracted through a Python script that communicated with the respective API. In turn, Facebook data was manually collected and coded. In the end, we identified 83 online events and 7017 social media posts.8 We must, of course, acknowledge that the effort necessary to post something on social media is not comparable to the effort put into organizing an on-site or online event or any attempt to enter the traditional media agenda. In order to find a balance that allows for a comparison of campaign actions of different nature, we assumed that the effort put in the preparation of contents for the social media is about ten times lower than the effort associated with other campaign activities. Therefore, in this analysis, each value of the social media category corresponds to blocks of 10 publications.

These four indicators allowed us to compute, for each candidate and, subsequently, for each election campaign under study, the relative weight (%) of digital vs. nondigital campaign efforts.

3.1.2. Online Competition

In order to identify patterns of normalization or equalization in online campaigning, we used data regarding the presence of candidates on Facebook and Twitter during the two months preceding the 2016 and 2021 presidential elections. Apart from activity, we were also interested in measuring impact, i.e., engagement, one of the major goals of social media political communication nowadays (Bene 2021). The activity index used in this article refers to the number of posts and tweets each candidate published, while the engagement variable reports the number of reactions (favorites on Twitter, and likes and specific reactions on Facebook),9 comments to, and shares of such contents.10 Activity indexes based on the number of posts are mainstream in this line of research (e.g., Lev-On and Haleva-Amir 2018; Bene 2021), while reactions and shares are often used to measure engagement (e.g., Samuel-Azran et al. 2015; Yang and Kim 2017; Lev-On and Haleva-Amir 2018; Bene 2021). These variables were then correlated with information on the candidates’ financial investment in the campaign and on their electoral prospects, two indicators commonly used to distinguish between stronger and weaker parties or candidates (e.g., Koc-Michalska et al. 2016; Yang and Kim 2017; Lev-On and Haleva-Amir 2018; Bene 2021). With regards to electoral prospects, instead of the usual measure based on the election results obtained in the previous race, which was unfeasible given that most candidates under analysis were running for the first time, we used poll data published in early January 2016/2021. Correlations are merely used to gauge the existence and magnitude of patterns of normalization (positive coefficients) or equalization (null or even negative coefficients) in terms of online competition.¹¹

3.1.3. Professionalization

The definition adopted here and our desire to interact with extant research on the country under study led us to use official campaign budgets and expense reports to measure campaign professionalization (in line with Lisi 2013, and Ferrinho Lopes et al. 2023), instead of candidate surveys or interviews (e.g., Gibson and Römmele 2009; Grusell and Nord 2020; Sampugnaro and Montemagno 2021). There are other studies of campaign professionalization that do look into campaign finances (e.g., Tenscher et al. 2012; Tenscher 2013; Mykkänen et al. 2021) but focus on the amount of money available to parties and candidates for the campaign and not on how that money is distributed among different categories of expenses.

In more concrete terms, we build on data on campaign finances available through the official documents that candidates deliver to the ECFP.12 We analyzed two types of spending data: campaign budgets (delivered up to one month before the election) and actual expenses (delivered up to six months after Election Day). Five categories of spending were considered: professionalization (campaign planning and design and hiring communication agencies and market research agencies, thus including all the expenses associated with campaign conception, development and overview), communication (which includes structures, posters, as well as printed and digital communication), on-site campaign (street campaigning, rallies, shows and whistle-blow tours, souvenirs for voters, and other gifts), administrative and operational costs (undifferentiated campaign staff, printing costs and preparation of official documents, as well as travel expenses, meals and campaign vehicles), and others. Our focus is, of course, placed on the first category, which constitutes a measure of the investment in the core features13 of a professionalized campaign (external public relations/media consultants, computerized databases, opinion polling, opposition research; Gibson and Römmele 2001, 2009).

4. Results

4.1. Digitalization

We start by testing the hypothesis that the pandemic context paved the way for a stronger bet on digital campaigning in 2021 vis-à-vis 2016 (H1a). The data presented in Figure 1 partially corroborates this assumption, showing that the 2021 election was characterized by a considerable investment in online events, when compared with 2016. Nonetheless, the relative weight of social media has not changed. In general terms, the 2021 campaign was more digital than the 2016 campaign due to the investment in online events but not to an increase in social media campaigning. Our first hypothesis (H1a) is therefore only partially confirmed. Interestingly enough, along with a foreseeable decrease of on-site campaign activities, we also note that legacy media presence increased considerably in relation to 2016, which indicates an effort of both the media and the candidates to use this channel in order to overcome pandemic-related campaign difficulties.

Looking only into the candidates running in both elections (to exclude the possibility of the patterns above being due to two different pools of candidates) reinforces the conclusion of a growing investment in the online sphere (Figure 2). This is particularly true in the cases of Marisa Matias and Vitorino Silva. First, in 2021 they promoted several online initiatives (corresponding to 26.1 and 11.5% of their activities, respectively), something that did not happen at all in 2016. Moreover, Vitorino Silva increased his investment in social media campaigning, going from 246 to 420 posts produced. Marisa Matias carried out a remarkably digital campaign in 2021: two-thirds of her campaign is grounded on online events and social media contents, being the most notorious difference vis-à-vis 2016 the investment in online events. On the contrary, Marcelo Rebelo de Sousa chose not to use digital tools on both occasions.14 Interestingly enough, both the incumbent president and Marisa Matias carried out less intense campaigns in 2021, as the total number of activities carried out shrunk; the opposite trend is observed for Vitorino Silva (Figure 2).

When we compare the pre- and post-lockdown periods of the 2021 campaign in search of an impact of a worsened pandemic situation on the behavior of the same pool of presidential candidates, the evidence leads us to reject our hypothesis. The data displayed in Figure 3 shows that whereas the relative weight of online events and presence in the legacy media increased after the lockdown, the clout of social media retracted considerably, while the role of on-site campaigning remained largely the same. Overall, this means that the campaign was even less digital after the declaration of lockdown than in the previous period. This seems to be a special feature of 2021, as in 2016 the differences between the pre-campaign and official campaign periods were less remarkable (and related, first and foremost, with an increase of the relative weight of on-site campaign activities; see Figure A1 in the Appendix A). The decreased degree of digitalization after the lockdown is first and foremost due to the lower intensity of social media use by André Ventura and Tiago Mayan Gonçalves after January 13, which, unlike the case of Marisa Matias or Vitorino Silva, was not compensated by an investment in online events (see Figure A2 in the Appendix A).

4.2. Online Competition

Figure 4 presents data on the intensity of social media presence (i.e., the number of published contents) and the average engagement (by post) in the 2016 and 2021 campaigns. On this figure, the candidates are organized from the most to the least voted.

This data seems to indicate that the normalization pattern, typical of the Portuguese case (see Santana-Pereira 2022) is, to a certain extent, observed in 2016, especially if one excludes a few outliers (Henrique Neto in terms of activity, Paulo Morais in terms of engagement, and Marcelo Rebelo de Sousa, who did not engage in online campaigning). Even if we include these candidates, Pearson correlations between activity and campaign expenses, between engagement and campaign expenses, and between engagement and electoral prospects are nonnegligible, although statistically nonsignificant due to the small N (r = 0.446, r = 0.394, and r = 0.396, respectively). In other words, the candidates with the highest financial resources were moderately more prone to online campaigning, and candidates with better electoral prospects were also more likely to engage efficiently with social media users. On the contrary, the correlation between activity and electoral prospects is negative (r = −0.326), which would suggest minor parties overcoming their offline shortcomings via intense online campaigning, i.e., equalization. This pattern is, however, clearly due to the main candidate refraining from using social media on his campaign, thus behaving as an outlier.

At first glance, the scenario did not change considerably in 2021, especially if one excludes Marcelo Rebelo de Sousa and the modest engagement capacity of Ana Gomes’s online campaign (Figure 4). Looking at the complete set of candidates, the correlation between campaign spending and social media activity is even stronger in 2021 (r = 0.667) than in 2016. Nevertheless, the correlation between electoral prospects and online activity is, again, negative, although of a stronger magnitude than in 2016 (r = −0.738) and reaching statistical significance with a 90% confidence level. In short, H2a, postulating that the patterns of normalization in terms of activity were less strong in 2021 than in 2016, is only partially accepted, as the pattern depends on the indicator of candidate strength used.

Regarding engagement, which increased in 2021 vis-à-vis 2016 (Figure 4), the pattern of normalization indeed tends to be weaker in the most recent election, as the correlations between engagement and campaign expenditures (r = 0.270) and, to a lesser extent, between engagement and electoral prospects (r = 0.374), find no statistical significance and are of lower magnitude in 2021 vis-à-vis 2016. This evidence thus allows us to accept H2b.

To better understand the role of the worsening pandemic scenario and lockdown, Figure 5 and Figure 6 display activity and engagement data for 2021 in two different periods: before and after the lockdown was declared. Since we are comparing two periods of different length15 using absolute values (and not relative values as in the previous section), we make them comparable by looking at the average number of daily publications and the average engagement by post. Overall, the lockdown period is marked by an increase in candidates’ activity and engagement, even though to different degrees (Figure 5 and Figure 6). However, the same pattern is also observed when one compares pre-campaign to official campaign periods of 2016 (Figure A3 and Figure A4 in the Appendix A).

The relationship between social media activity and campaign expenses was somewhat stronger in the pre-lockdown (r = 0.647) than in the post-lockdown period (r = 0.555), although none of the Pearson coefficients are statistically significant. In addition, the negative relationship between activity and electoral prospects was weaker before (r = −0.574) than after the lockdown (r = −0.785), with this latter coefficient being significant with a 95% confidence level. In sum, this evidence points to a milder pattern of normalization of online competition in terms of bulk of activity after the lockdown, as expected. This conclusion is further strengthened if one compares the relationship between candidate strength and social media activity during the pre-campaign and campaign periods of 2016, as there was no significant change in the magnitude of that relationship (Pearson coefficients of 0.4). There was, however, an increase in the magnitude of the correlation between electoral prospects and online campaign intensity from the pre-campaign (r = −0.234) to the official campaign (r = −0.457) periods in 2016. Yet, those coefficients fail to reach statistical significance. This does not allow us to establish that in 2021 the strengthened pattern of equalization (when candidates are ranked based on their electoral appeal) after the lockdown is due to this event and not to the fact that candidates were entering a different stage of the campaign cycle. Hypothesis H2c is therefore only partially confirmed.

In terms of engagement (Figure 6), the normalization pattern is indeed stronger after the lockdown: while the relationship with campaign expenses is negligible in both moments, the positive correlation between engagement and electoral prospects is weaker before (r = 0.451) than after (r = 0.728) the lockdown. In 2016, there were no significant shifts in coefficient magnitude between pre-campaign and official campaign periods (all equally low, close to zero), which means that this strengthened normalization pattern is likely not due to the beginning of a more official campaign period but to the post-lockdown context. This disconfirms our expectation that normalization patterns after the lockdown would be milder in terms of engagement (H2d).

4.3. Professionalization

We hypothesized that a higher proportion of campaign finances was allocated to professionalization in 2021 as compared to 2016 (H3a). Before having a look at Figure 7, which compares the campaign investment focus in the 2016 and 2021 presidential election campaigns, we must underline the fact that in 2021 campaign expenses retracted to less than a third compared to the previous presidential contest (from around 3.18 to 1.05 million euros). This was not due to the lower number of candidates in 2021 vis-à-vis 2016, as the average campaign expenses by candidate also declined from around 318,000 to 150,000 euros. In turn, this decrease might be due to candidates feeling they should not spend too much money under an environment characterized by the economic wounds caused by the pandemic, as well as to a trend of decreasing spending in presidential campaigns observed from 2006 onwards (Serra-Silva and Santos 2022). Of course, less money in absolute terms lowers the odds of professionalized campaign planning and development becoming more important in relation to other traditionally strong areas of campaign investment (Gibson and Römmele 2001, 2009).

Interestingly, the relative weight of professionalization was fairly similar in 2016 and 2021: 10.5 and 9.9% of total expenses, respectively (Figure 7). In turn, communication experienced a modest growth (3.7 points), while administrative costs increased considerably. There was also, unsurprisingly, a lower weight of investment in on-site campaign events in 2021. These results disconfirm H3a, suggesting that the pandemic did not lead to increased spending in professionalization.

Since professionalization investment decisions can be impacted by individual or party preferences (Gibson and Römmele 2001, 2009), and that the two scenarios compared above were composed of campaigns carried out by different people with distinct party links, we now shift to an analysis of the investment focus of the three candidates running in both elections (Figure 8). This analysis reinforces the previous conclusions. On the one hand, only Marcelo Rebelo de Sousa invested in professionalization. The fact that Marisa Matias did not spend money on professionalization can be explained by her leftwing ideological stance (Gibson and Römmele 2001, 2009)—in fact, both the BE candidate and the Communist/Green candidates in 2016 and 2021 completely refrained from investing in professionalization (

Table A1. Planned investment and actual expenses in the 2021 Presidential election campaigns.

	Professionalization		Communication		Events		Administrative and Operational Costs		Others		Total
	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Change
Marcelo Rebelo de Sousa	1500 (6%)	15,276 (61.2%)	3500 (14%)	1047 (4.2%)	0	0	16,000 (64%)	7504 (30.1%)	4000 (16%)	1100 (4.4%)	25,000	24,927	−73
Ana Gomes	17,500 (32.7%)	44,477 (32.8%)	17,000 (31.7%)	25,573 (18.9%)	8500 (15.9%)	31,051 (22.9%)	5000 (9.4%)	34,352 (25.4%)	5500 (10.3%)	0	53,500	135,453	+81,953
André Ventura	25,000 (15.6%)	36,900 (18.4%)	75,000 (46.8%)	56,734 (28.3%)	40,000 (25%)	6123 (3%)	10,000 (6.3%)	21,013 (10.4%)	10,000 (6.3%)	80,343 (39.9%)	160,000	201,112	+41,112
João Ferreira	0	0	225,000 (50%)	110,921 (40.4%)	35,000 (7.8%)	24,705 (9%)	170,000 (37.8%)	138,637 (50.6%)	20,000 (4.4%)	0	450,000	274,264	−175,736
Marisa Matias	0	0	155,473 (60.6%)	145,514 (41%)	52,466 (20.3%)	121,079 (33.7%)	48,679 (19%)	91,279 (25.5%)	0	0	256,618	357,872	+101,255
Tiago Mayan Gonçalves	10,450 (27.2%)	6765 (14.3%)	18,000 (46.8%)	27,204 (57.5%)	3000 (7.8%)	8194 (17.4%)	5500 (14.3%)	5120 (10.8%)	1500 (3.9%)	0	38,450	47,284	+8834
Vitorino Silva	3000 (18.7%)	0	6000 (37.5%)	3000 (41.9%)	0	0	1000 (6.3%)	1655 (23.2%)	6000 (37.5%)	2500 (34.9%)	16,000	7155	−8845
Total	57,450 (5.8%)	103,418 (9.9%)	499,973 (50%)	369.993 (35.3%)	138,966 (13.9%)	191,152 (18.2%)	256,179 (25.6%)	299,560 (28.6%)	47,000 (4.7%)	83,943 (8%)	999,568 (100%)	1,048,067 (100%)	+48,499 (4.9%)

Source: Own coding of data collected from ECFP (https://www.tribunalconstitucional.pt/tc/contas_eleicoes-pr.html, accessed on 11 September 2022).

and

Table A2. Planned investment and actual expenses in the 2016 Presidential election campaigns.

	Professionalization		Communication		Events		Administrative and Operational Costs		Others		Total
	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Budget	Expenses	Change
Marcelo Rebelo de Sousa	20,000 (12.7%)	114,376 (63.6%)	55,000 (35.2%)	12,121 (6.8%)	20,000 (12.7%)	25,945 (14.5%)	40,000 (25.4%)	13,045 (7.3%)	22,000 (14%)	14,019 (7.8%)	157,179	179,507	+22,328
António Sampaio da Nóvoa	50,000 (6.7%)	46,475 (5%)	240,000 (32.3%)	345,776 (37.4%)	120,000 (16.2%)	354,146 (38.3%)	200,000 (27%)	150,937 (16.3%)	132,000 (17.8%)	27,157 (3%)	742,000	924,493	+182,493
Marisa Matias	0	0	128,301 (28.2%)	78,709 (13.1%)	143,910 (31.7%)	125,810 (20.7%)	47,448 (10.4%)	98,649 (16.2%)	135,000 (2.7%)	298,545 (49.1%)	454,659	607,715	+153,056
Maria de Belém Roseira	84,271 (13%)	56,478 (10.4%)	195,877 (30.1%)	183,634 (33.9%)	267,968 (41.2%)	236,687 (43.7%)	99,613 (15.3%)	60,716 (11.2%)	2268 (0.4%)	4380 (0.8%)	650,000	541,896	−108,104
Edgar Silva	0	0	400,000 (53.3%)	276,281 (47.5%)	185,000 (24.7%)	144,659 (24.9%)	145,000 (19.3%)	159,189 (27.4%)	20,000 (2.7%)	984 (0.2%)	750,000	581,114	+168,886
Vitorino Silva	2000 (4%)	0	13,000 (26%)	0	9000 (18%)	0	10,000 (20%)	926 (11.5%)	16,000 (32%)	7229 (88.5%)	50,000	8159	+41,841
Paulo de Morais	13,714 (23%)	13,714 (23%)	16,716 (28%)	16,716 (28%)	6423 (10.8%)	6423 (10.8%)	22,684 (38%)	22,684 (38%)	0	0	59,539	59,539	=
Henrique Neto	14,000 (5.1%)	77,509 (31.2%)	67,000 (24.3%)	91,862 (39.9%)	28,000 (10.2%)	45,106 (18.1%)	140,000 (50.9%)	31,288 (12.6%)	26,000 (9.5%)	3005 (1.2%)	275,000	248,771	+26,226
Jorge Sequeira	15,000 (12.1%)	3690 (61.2%)	43,000 (34.8%)	425 (7.1%)	13,000 (10.5%)	270 (4.5%)	8500 (6.9%)	40 (0.7%)	44,000 (35.6%)	1600 (26.6%)	123,500	6026	+117,474
Cândido Ferreira	30,000 (50%)	23,207 (75.8%)	14,000 (23.3%)	0	6000 (10%)	5456 (17.9%)	4000 (6.7%)	0	6000 (6.7%)	1968 (6.4%)	60,000	30,632	−29,368
Total	228,986 (6.9%)	335,451 (10.5%)	1,172,895 (35.3%)	1,005,527 (31.6%)	799,302 (24%)	944,506 (29.7%)	717,246 (21.5%)	537,482 (16.8%)	403,268 (12.2%)	358,890 (11.3%)	3,321,698 (100%)	3,181,858 (100%)	−139,840 (−4.3%)

Source: Own coding of data collected from ECFP (https://www.tribunalconstitucional.pt/tc/contas_eleicoes-pr.html, accessed on 11 September 2022).

in the Appendix A). Regarding Vitorino Silva, the absence of investment in professionalization may be first and foremost due to lack of financial resources—on both elections, Silva spent less than 10,000 euros (Table A1 and Table A2 in the Appendix A). On the other hand, there was no change in the incumbent’s relative investment in professionalization in 2016 and 2021. On both occasions, his investment was considerable in proportion to other spending areas—something we would expect from a leading candidate with a catch-all stance and ideologically closer to the right (Gibson and Römmele 2001, 2009; Mykkänen et al. 2021). The investment was, nevertheless, very modest in absolute terms in 2021, as Marcelo Rebelo de Sousa considerably reduced campaign costs vis-à-vis 2016 (from around 180,000 euros to around 25,000 euros; see Table A1 in the Appendix A).

To conclude, we look into the influence of the pandemic by erasing the possibly confounding impact of time passing by, as evidence points to a growing trend of professionalization of campaigns in Portugal and elsewhere over the years (Lisi 2013; Tenscher 2013; Grusell and Nord 2020). Figure 9 allows for a comparison between the planned investment (budgets) and the actual investment (actual expenses communicated after the elections) for both the 2016 and 2021 elections. This data shows the relative weight of expenses on professionalization is 4.2 points above from what had been planned in 2021. This is mainly due to Marcelo Rebelo de Sousa spending ten times more than initially planned (Table A1 in the Appendix A). To the contrary, communication retracted from 50 to 35.3 percent. There was also an increase of the relative weight of the investment in on-site campaign activities. In order to find out if this was due to the pandemic or an effect of usual trends of change between budgets and expenses, we carried out the same comparison for 2016. We discovered that a similar phenomenon happened in that context: investment in professionalization also grew in relation to what had been budgeted. Therefore, COVID-19 and the lockdown cannot be said to have had an impact in boosting investment in professionalization in 2021. Hypothesis H3c is thus discarded.

5. Conclusions

This article assessed the impact of the COVID-19 pandemic in campaign strategies, using a most similar systems design to compare two presidential campaigns in Portugal, as well as the pre- and post-lockdown periods of the most recent campaign. The propriety of interest is the pandemic context of the 2021 election, which created pitfalls to candidates willing to mobilize and persuade voters. Our expectation was that candidates curbed the hazards brought about by the pandemic through a stronger investment in digitalization and professionalization. The evidence, however, points to a different panorama.

First, our data shows that the 2021 Portuguese presidential elections did not represent a turning point in terms of transition from traditional to digital campaigning in Portugal, as early accounts of the campaign suggested (e.g., Serra-Silva and Santos 2022). In 2021, online events were more frequent, but social media activity was not relatively more important than in 2016. In this regard, two-times candidate Vitorino Silva is an exception, as he augmented both his use of social media and online events between 2016 and 2021. Moreover, a closer glance at the 2021 elections reveals that despite the mandatory confinement declared on January 13, on-site campaigning did not retract when compared to the period before the lockdown, while digital campaigning did become proportionally less relevant. The most important impact of the lockdown was an increase in the relative weight of the legacy media on the candidates’ agendas. This might be because candidates felt that since people spent more time at home during the day, this could have led to increased media consumption. In short, while the 2021 campaign was not extravagantly more digital than the previous one, especially if one disregards online events (whose long-term permanence in the campaign toolbox is doubtful), the post-lockdown campaign was even less digital.

Second, when looking at the 2021 candidates’ use of social media, we were not able to find a clearcut change in terms of online competition patterns vis-à-vis the 2016 election. Our mixed evidence suggests that the indicator chosen to measure candidate strength is particularly relevant in contexts in which the most popular candidate is not the one spending more money on the campaign, as it is the case of both the 2016 and the 2021 elections. Additionally, the differences in our conclusions vis-à-vis Ferrinho Lopes et al. (2023) highlight the importance of considering broader campaign periods and distinguishing between different stages of the campaign. However, not all results are inconsistent with our expectations—for instance, we spotted a shift towards a less flamboyant pattern of normalization in terms of impact in 2021 vis-à-vis 2016, with differences in engagement levels between candidates being less strongly related with their spending or poll results.

Third, the COVID-19 pandemic did not seem to have boosted professionalization, as there are no differences between 2016 and 2021. In this context, the ideological component of the professionalization investment decision is worth mentioning: from the three candidates running to both elections, only Marcelo Rebelo de Sousa—the incumbent, leading, center-right candidate—invested in professionalization in both elections. A second relevant pattern stemming from the comparison between the campaign budgets and the final campaign expenses is that investment in on-site campaign activities increased proportionally in the later stages of the campaign both in 2016 and 2021. Again, no lockdown impacts could be traced.

Altogether, the surveyed evidence leads to the conclusion that the 2021 campaign was far from been a watershed moment in terms of campaign strategies put forward by Portuguese candidates. These findings have important implications for the understanding of how and when COVID-19 becomes normalized in politics: as time passes by, regardless of massive infections and deaths by COVID-19, candidates appear to be unwilling to act differently. This scenario can be explained by a sort of flight to safety (Bisbee and Honig 2022) by candidates themselves—in exceptional circumstances, sticking with what is known and has been done before might be the best strategy. In short, in 2021 presidential candidates chose to sail the COVID-19 pandemic-related uncharted waters with old, reliable-looking boats. Amongst the incentives that might have worked against candidates deciding to act differently, we must underline the lack of financial resources: the budgets in 2021 were remarkably modest, which, of course, hinders the degree of professionalization campaigns can aim to reach (but not digitalization, as online campaigning is often regarded as a low-cost way of campaigning, cf. Serra-Silva and Santos 2022). The belief that there were more important things to focus on (namely the discussion on the combat against COVID-19, highly salient during the campaign; Serra-Silva and Santos 2022) than revolutionizing one’s presence in an election campaign with a very unfortunate timing might also have played a role. Lastly, the fact that these were second-order and uncompetitive elections might have had an effect that was opposite to the expected: instead of representing a safe ground to do things differently, it demotivated candidates from trying new ways of campaigning.

When we keep doing the same thing, we should not expect different results. Indeed, the stability of the campaign strategies put forward by presidential candidates in 2021 when compared to 2016 can perhaps be linked to the absolute lack of surprise in terms of electoral results: the incumbent candidate won by a landslide as predicted by pre-electoral polls, while the results achieved by the other candidates were also similar to their vote share forecasts. Moreover, we can also connect it with the inability of combatting COVID-19-related abstention by engaging and mobilizing voters. Indeed, these were the less participated presidential elections in the history of the country (turnout = 39%), and there is evidence that fear of contracting COVID-19 was a key factor on the decision to abstain (Dias 2022). In a nutshell, the business-as-usual presidential campaigns of 2021 were seemingly not able to win neither the battle of incumbency advantage nor pandemic-related demobilization.

This article focused on a wide array of objective data on how the 2021 (and the 2016, for comparative purposes) presidential election campaigns were carried out, namely in terms of investment in professionalization and digitalization. It was not possible, however, to enrich the analysis with subjective accounts derived from candidate and campaign staff surveys (nonexistent) or interviews (unfeasible within the time frame for this research). Future research on this matter should therefore try to complement objective and subjective measures of digitalization and professionalization.