Introduction
The constitutional strength of a presidency largely defines the role of the president in each system. Whereas in some systems the executive authority is in the hands of the president, in other systems, power is divided between the president and the government. The existence of a prime minister usually weakens the president at least to some extent but may also work in favor of the president’s popularity ratings. Yet we have no systematic evidence of the connection between presidential powers and presidential popularity across regions and different constitutional settings. Moreover, the majority of studies explaining presidential popularity have focused only on presidential countries, such as the USA and countries in Latin America, where the presidents possess very similar powers, while other countries and regime types with constitutionally weaker presidents have remained understudied, not least in Europe.
This study addresses this major research gap by examining the aggregate-level patterns of presidential popularity across semi-presidential and presidential regimes with presidential powers as the main explanatory variable. The aim is to contribute to the literature on public opinion on presidents from a new perspective by showing empirical evidence of the association of presidential popularity with presidential powers in different constitutional settings. The research question is ‘Does the level of presidential powers explain differences in public opinion on presidents?’.
The focus is on semi-presidential and presidential regimes as they share the same features of separate origin (separate election of the president and the legislature) and separate survival (the president serves for a fixed term and cannot be removed by the assembly majority) (Samuels and Shugart, Reference Samuels and Shugart.2010), yet the power-sharing inside the executive branch largely distinguishes the former from the latter regime type. Due to this profound difference, presidents in presidential regimes tend to be more powerful than presidents in semi-presidential regimes (e.g., Siaroff, Reference Siaroff2003; Elgie, Reference Elgie2018). The basic premise here is that the division of powers between the major political institutions should impact peoples’ expectations and perceptions of the president. At the same time, the constitutional strength of the presidents varies within the regime types as well, especially within semi-presidentialism. Therefore, if we want to understand the dynamics of public opinion on presidents from a comparative perspective, the diversity of presidential powers should be considered.
Theoretically, this study relies on responsibility theory used often in the literature on presidential popularity (approval). It suggests that people judge incumbent officeholders, for example, for changes in the economy if they hold them accountable for those matters (e.g., Nannestad and Paldam, Reference Nannestad and Paldam.1994; Stegmaier et al., Reference Stegmaier, Lewis-Beck, Park., Arzheimer, Evans and Lewis-Beck2017). In the context of this study, the effect should be more evident among presidents with stronger constitutional powers. Empirically, it provides a multi-level analysis of the impact of presidential powers on public opinion on presidents in 15 countries across semi-presidential and presidential regimes between 2000 and 2020. It combines public opinion data from the Executive Approval Project by Carlin et al. (Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2019), presidential power scores from Siaroff (Reference Siaroff2003), and other contextual factors from economic development to electoral cycle.
The study confirms that weaker presidential powers are associated with more positive opinions toward the presidents, reflected both in terms of the general level of presidential popularity between weak and strong presidents and in the extent to which the honeymoon period, changes in unemployment rate, and GDP per capita growth explain whether people ‘reward’ or ‘punish’ the presidents. Moreover, constitutionally more powerful presidents tend to be less popular in general and more likely to be criticised for the worsening economy and later in the electoral cycle, confirming the expected responsibility thesis. These findings underline the importance of including presidential powers as an explanatory variable when studying public opinion on presidents across countries.
In next section, I present the theoretical background of the study by discussing regime-level differences in terms of presidential powers and the role of the president, as well as explanations of public opinion on presidents in the context of earlier studies before summarising the main hypotheses for the subsequent analysis. Then, follows a section introducing data and the methods before the results of the empirical analysis. Finally, I discuss the empirical findings with some reflections about implications for future research.
Theoretical framework
Presidential powers and regime-level differences
All presidents carry certain expectations. Usually, presidents are expected to act as guardians of ‘national unity’ and to represent the country abroad (Tavits, Reference Tavits2009; Duvold and Sedelius, Reference Duvold and Sedelius.2022). The actual role of the president does, however, vary between countries and regime types. Some presidents have considerable powers and some have more of a ceremonial role. Presidents in presidential regimes are by and large the main political leaders of the countries with strong constitutional powers, while presidents in semi-presidential regimes need to balance between representing the nation and cooperating with prime ministers. This key difference between the systems should work in favor of presidents in the latter regime type as they usually face less expectations in terms of domestic politics.
By definition, semi-presidentialism is a system ‘where there is both a directly elected fixed-term president and a prime minister and cabinet who are collectively responsible to the legislature’ (Elgie, Reference Elgie2011: 3). In such systems, the presidents are, sometimes even by constitutional rules, expected to stay above party politics, which might account for some of their higher popularity ratings. They may be experienced politicians but still able to distance themselves from party politics in the eyes of the citizens after being elected. Prime ministers, who usually possess more powers but are systematically less popular, should in turn suffer more from their position as they cannot escape responsibility for uncomfortable decisions in daily national politics. (See, for example, Tavits, Reference Tavits2009; Raunio and Sedelius, Reference Raunio and Sedelius.2020.)
Whereas presidents in semi-presidential regimes may avoid party-political battles, presidents in presidential regimes are usually highly partisan. For example, in the context of USA presidents, Cavari (2017) argued that both presidents and citizens operate in a partisan environment, which impacts not only the actions of the presidents but also public opinion. Cavari (2017: 21) summarised that a ‘president’s actions are labeled and understood in partisan terms, and the effect of these actions on public opinion is mediated and enhanced by partisan labels’. People, for example, tend to reward and punish the presidents for the country’s economic performance especially if the president represents the opposite party (Lebo and Cassino, Reference Lebo and Cassino.2007).
Following the definition by Elgie (Reference Elgie2011), semi-presidentialism is currently the most common regime type in Europe, covering a wide range of constitutions from very powerful to very weak presidencies (Anckar, Reference Anckar2022).Footnote 1 As an example, both Finland and France can be categorized as semi-presidential, yet whereas the powers of the Finnish president are limited to foreign and security policy, the French president largely dominates both foreign and domestic policy issues. Within the regime type, France represents one of the strongest presidencies (e.g., Siaroff, Reference Siaroff2003; Doyle and Elgie, Reference Doyle and Elgie.2016; Elgie, Reference Elgie2018), and support for the French presidents has been relatively low in the early twenty-first century. In contrast, Finland represents one of the weakest presidencies within the semi-presidential category (ibid.). After the constitutional reform at the turn of the twenty-first century, Finnish presidents have been vested with very weak legislative powers yet with a substantial role in foreign policy. Since this constitutional reform, the Finnish presidents have retained very high popular ratings, where the two most recent presidents, Sauli Niinistö and Tarja Halonen, have received approval ratings well above 80 percent for most of their incumbency.Footnote 2
These examples indicate that it is not the regime type itself that determines the level of presidential popularity, but the division of powers in a semi-presidential constitution that enables the president to stand out from other political institutions in a way that benefits the president. Simultaneously, the complexity of the role of the presidents should be acknowledged as not all presidents use their constitutional prerogatives, and in some cases, despite their weak powers the presidents can also become very powerful.
Popular elections should also bring certain nuances to the nature of the presidential office in both presidential and semi-presidential regimes. It has for example been argued that directly or popularly elected presidents can often act more actively and independently from parliament and government (Köker, Reference Köker2017). Moreover, due to the mode of the elections, the presidents should by default be more connected to the public than prime ministers or indirectly elected presidents. This connects to the common assumption that presidents who are popularly elected might justify their actions with the direct mandate given by the public (Duverger, Reference Duverger1980; Baylis, Reference Baylis1996; Lijphart, Reference Lijphart1999). And if they are popular too, with this ‘legitimacy’ they might push the limits of their constitutional prerogatives to impose their policy agenda. Constitutionally weaker presidents might, for example, avoid publicity in tricky situations or act more boldly through informal channels to ‘compensate their limited constitutional powers’ (Raunio and Sedelius, Reference Raunio and Sedelius.2020: 29). In semi-presidential regimes, this has even led to intra-executive conflicts between the presidents and prime ministers, especially in Central and Eastern Europe (e.g., Protsyk, Reference Protsyk2005, Reference Protsyk2006; Sedelius and Mashtaler, Reference Sedelius and Mashtaler.2013; Sedelius and Ekman, Reference Sedelius and Ekman.2010; Elgie, Reference Elgie2018). Higher popularity ratings are important for stronger presidents in presidential regimes as well, as they might want to use the ‘going public’ strategies to promote their policies, especially if they are relatively more popular than other key political actors (Kernell, Reference Kernell2007; Ponder, Reference Ponder2018).
The influential position of the president might also turn against them if they act too boldly or decide not to act when expected. For example, recent studies have shown that the COVID − 19 pandemic boosted some presidents’ popularity ratings, but not without exception. For example, studying the popularity of four Latin American presidents in the early stages of the pandemic, Sosa-Villagarcia and Lozada (Reference Sosa-Villagarcia and Lozada.2021) noted that the popularity ratings of the Peruvian president increased quite rapidly, but presidents in Chile, Brazil, and Mexico faced different effects. The authors linked these changes in the popularity ratings to three factors: ‘timing and stringency of the measures, policy framing, and the role of the opposition (Sosa-Villagarcia and Lozada, Reference Sosa-Villagarcia and Lozada.2021: 87). This was well reflected in the actions of Jair Bolsonaro, the president of Brazil, whose reactions were, compared to many other heads of state, very controversial and denialist toward the virus (Lasco, Reference Lasco2020: 1420), which supposedly harmed his popularity ratings. From another point of view, the popularity of the Lithuanian president Gitanas Nauseda decreased at the beginning of the pandemic despite his limited powers in domestic politics. Park (Reference Park2020) connected the drop in Nauseda’s popularity ratings to his lack of leadership and limited number of appearances in public. After this backlash, Nauseda tried to raise his profile in domestic politics by introducing several bills to the parliament (Park, Reference Park2020). In Finland, on the other hand, President Sauli Niinistö remained very popular despite (or because of) his limited role in the crisis situation (Kujanen and Raunio, Reference Kujanen and Raunio.2020).
Responsibility theory and other explanations of presidential popularity
Presidential popularity is a complex phenomenon with different contextual factors explaining the dynamics of public opinion on presidents in different systems. For example, Donald Trump’s approval ratings were relatively low yet very stable and more polarised when compared to other USA presidents, and common factors such as the economy and major events did not seem to have the same impact on his popularity ratings – reflecting his ‘unusual’ presidency (Newman and Otto, Reference Newman and Otto.2022). As another example, Emmanuel Macron’s popularity ratings have followed the usual path with a peak of popularity at the beginning of his term, followed by a decline and then a rise before the 2022 presidential elections. Otherwise, his popularity ratings have suffered from ‘his’ government’s policies and the so-called ‘yellow vest movement’, but on the other hand, he has enjoyed quite strong support too especially at the beginning of the COVID-19 pandemic (Clegg, Reference Clegg2022).
Scholars have tried to explain public opinion on presidents for decades, yet almost all empirical studies focus on presidential regimes, such as the USA or Latin American countries. Some studies focus on French presidents (e.g., Lewis-Beck, Reference Lewis-Beck1980; Lafay, Reference Lafay, Eulau and Lewis-Beck1985; Conley, Reference Conley2006; Turgeon et al., Reference Turgeon, Bélanger and Nadeau.2015), yet, otherwise, there seems to be no similar studies on other semi-presidential countries. Comparative studies of public opinion on presidents are also rare and they mainly focus on presidential countries (e.g., Carlin et al., Reference Carlin, Love and Martínez-Gallardo.2012a, Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2018; Jung and Oh, Reference Jung and Oh.2020). As an exception, Carlin et al. (Reference Carlin, Martínez-Gallardo and Hartlyn2012b) broadened the comparison to include parliamentary regimes, yet their study does not focus solely on presidents and excludes semi-presidential regimes entirely.
Many studies follow the seminal work of Mueller (Reference Mueller1970) and have, for example, found that international events and crises, the electoral cycle, and changes in the economy have an impact on presidential popularity ratings. The first refers to the so-called rally around the flag effect, a situation where people start ‘rallying’ behind their leaders in times of crisis. Regarding the electoral cycle, it seems that presidents enjoy higher popularity during the first months in office, which is often referred to as the honeymoon effect, and that there may be a ‘backswing’ at the end of the president’s term during the last few months before the next elections. Overall, however, the popularity ratings fade with time, which in turn is called the cost of ruling effect (e.g., Magalhães et al., Reference Magalhães and Andeweg2020; Erikson, Reference Erikson2012; Berlemann and Enkelmann, Reference Berlemann and Enkelmann.2014; Carlin et al., Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2018). As explained by Lewis-Beck and Stegmaier (Reference Lewis-Beck and Stegmaier.2013: 379): ‘The idea is that governments pay a cost for being in office. Over time, they strike bargains and create enemies, and their supporters become disillusioned and withdraw’.
Much of the research has focused on the third aspect – economic development. These studies suggest that macroeconomic factors, such as unemployment, inflation, and GDP per capita growth, drive presidential popularity (e.g., Magalhães et al., Reference Magalhães and Andeweg2020; Nannestad and Paldam, Reference Nannestad and Paldam.1994; Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier.2013; Berlemann and Enkelmann, Reference Berlemann and Enkelmann.2014; Stegmaier et al., Reference Stegmaier, Lewis-Beck, Park., Arzheimer, Evans and Lewis-Beck2017). This connects to the responsibility theory often referred to in presidential popularity studies as part of the vote and popularity functions (Nannestad and Paldam, Reference Nannestad and Paldam.1994; Lewis-Beck and Stegmaier, Reference Lewis-Beck and Stegmaier.2013; Stegmaier et al., Reference Stegmaier, Lewis-Beck, Park., Arzheimer, Evans and Lewis-Beck2017). In a nutshell, it has been argued that short-term changes in the political environment, such as the state of the economy, affect the popularity of the incumbent officeholders because people hold them responsible for these matters. This so-called ‘responsibility hypothesis’ should work only if the presidents have actual control over the matters, i.e., in countries where the presidents possess considerable powers in domestic politics (Nannestad and Paldam, Reference Nannestad and Paldam.1994), and where the ‘clarity of responsibility’ holds (Ningham and Whitten, 1993).
The responsibility theory thus suggests that in terms of public opinion, it should matter who is responsible for political decisions. Therefore, the constitutional powers of the political actors should be considered. Carlin and Singh (Reference Carlin and Singh.2015) adopted this idea by examining the association between presidents’ legislative powers and presidential popularity in 18 Latin American countries. Their results suggest that presidents who possess stronger legislative powers, and especially those who use them more frequently, face higher levels of economic accountability, which benefits the popularity ratings of the presidents during a better economic situation but turns into more critical views among the public under an economic downturn. In a similar vein, Elgie (Reference Elgie2018) found that presidential powers had an impact on economic voting, i.e., to what extent the economic development accounts for support for political institutions, with the effect of economic development being stronger in systems with stronger presidents. According to the results by Elgie (Reference Elgie2018), the effect was more evident under semi-presidentialism than under parliamentarism as the level of presidential powers varied more within the former regime type. Although Elgie (Reference Elgie2018) concentrated on the percentage of votes for the president’s party in presidential and parliamentary elections, the same logic should apply in the context of presidential popularity as vote and popularity functions are connected in the same theoretical framework.
In terms of economic voting, it has been found that the clarity of responsibility is more blurred in multiparty systems with coalition governments, while single-party governments and two-party systems may sharpen it (Nannestad and Paldam, Reference Nannestad and Paldam.1994; Stegmaier et al., Reference Stegmaier, Lewis-Beck, Park., Arzheimer, Evans and Lewis-Beck2017). In general, the clarity of responsibility should hold in presidential systems because executive power is not divided and thus the president and the president’s party are typically held responsible for the state of the country. In contrast, the dual-executive nature of semi-presidential regimes may cause confusion among the public about who is responsible. As a consequence, the public might sometimes prefer a more neutral appearance of the president as well as actions that push the constitutional prerogatives of the office.
One important implication of the previous studies is that institutional structures matter when people are evaluating the performance of the presidents. Whether or not this is based on citizens’ knowledge of the powers associated with each institution or other beliefs, weaker presidents appear as winners in both situations – they have fewer responsibilities in the constitution and more freedom to avoid difficult decisions. Prime ministers or presidents with more significant powers do not have such freedom. Of course, this does not mean that weaker presidents would be immune to changes in the political environment or other contextual factors, such as international events, but rather that more powerful presidents should be more sensitive to them, as the theory and previous empirical evidence suggests. In other words, the position of weaker presidents ‘above party politics’ should protect them to some extent. Given this, it might be that public support (or the lack of it) for constitutionally stronger presidents is more direct or ‘specific’, and for constitutionally weaker ones more indirect or ‘diffuse’, referring to the classical distinction of political support by Easton (Reference Easton1975).
Hypotheses
To sum up, I expect the constitutional constraints of presidents to impact citizens’ perceptions of them. This should be reflected not only in the general level of the popularity ratings between weak and strong presidents but also in terms of factors that cause variation in the popularity ratings. Recognising the constitutional differences between the systems and drawing on the earlier literature of the explanations of presidential popularity, I expect weaker presidential powers to result in higher and more stable popularity ratings that are not as much contingent on the national economic development or the passage of time in office. To test this argument, I formulate three hypotheses.
First, in terms of the general level of presidential popularity, I expect presidents with greater powers and responsibilities to be judged more critically than presidents who get to ‘stay above’ daily politics. Thus, presidents with weaker powers should enjoy relatively higher popularity ratings than presidents with stronger constitutional powers. This leads to the first hypothesis:
H1: Presidential popularity is positively associated with presidents’ weaker constitutional powers.
Second, I expect weaker presidents’ popularity ratings to be relatively more stable and thereby to vary less in terms of the electoral cycle. More specifically, in the spirit of the cost of ruling theory, constitutionally more powerful presidents’ popularity ratings should fluctuate and decrease more over time as they are, as the head executives, more involved in daily politics and there are therefore more opportunities for the public to be disappointed in them. Thus, the second hypothesis is:
H2: Popularity ratings of constitutionally stronger presidents fluctuate more over time than those of constitutionally weaker presidents.
Third, I expect the popularity ratings of presidents with stronger constitutional powers to be relatively more sensitive to national economic development. This assumption stems from the responsibility theory suggesting that short-term changes in the economy impact public opinion on presidents because they are held responsible for economic issues. However, if the constitutional powers of the president are weak and co-shared with the prime minister who is actually responsible for the domestic matters, the responsibility theory should not apply as it is likely that the citizens hold the prime minister accountable for economic policy and not the president. The third hypothesis is:
H3: Changes in the economy have stronger impact on public opinion on presidents with stronger constitutional powers than on presidents with weaker constitutional powers.
Data and Methods
The analysis covers 15 countries, eight presidential and seven semi-presidential, between 2000 and 2020 and includes altogether 68 presidents. Only countries categorized as democracies by the Polity5 project (Marshall and Gurr, Reference Marshall and Gurr2020) and rated as Free by Freedom House (2022) were included in the analysis, with a notion that the scores have varied over time and for some countries they have changed significantly during the first two decades of the twenty-first century. The indices measure democracy differently, yet they are both used by many scholars and correlated with each other (e.g., Högström, Reference Högström2013; Paldam, 2021). Otherwise, the selection of countries was mainly based on data availability from the Executive Approval Project (EAP) (Carlin et al., Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2019).Footnote 3 Moreover, countries that can be categorized as presidential or semi-presidential and had presidential popularity ratings separately in the EAP dataset were included in the analysis. Only Finland was not part of the dataset, but it was included due to the diversity of powers and data availability. Among semi-presidential countries, only Peru represents president-parliamentarism while the others represent premier-presidentialism.Footnote 4 The selection of the period from 2000 to 2020 (although in most countries until 2018 or 2019) for both popularity and economic time series was based on data availability. The regime types, countries, years, and number of presidents in each country are listed in Table 1.
List of countries and time periods in the dataset
For the Czech Republic, years 2000–2012 were excluded due to indirect presidential elections until 2012.
The analysis consists of descriptive figures and statistics, fixed effects regression models with interaction terms, and additional robustness checks. The descriptive figures show the general variation of public opinion on presidents in relation to the level of presidential powers. Basic mean comparisons (ANOVA and t-test) were performed to confirm whether there are statistically significant differences between the analysed groups. The main analysis consists of three regression models with quarterly data. Regression coefficients are applied to estimate the effect of different possible determinants of presidential popularity across different constitutions with special attention on the individual effect of presidential powers and conditional effects of the economic variables and electoral cycle. Additional robustness checks ensure that the models are not biased by the choice of the specific presidential power scores (Siaroff, Reference Siaroff2003) or the structure of the data.
The dependent variable is the percentage of presidential popularity, measuring people’s evaluations of the performance of the president and trust in the president. More specifically, the variable consists of various opinion polls for each country. The opinion data is from the Executive Approval Dataset (EAD) by Carlin et al. (Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2019), supplemented with opinion polls collected from Finland.Footnote 5 The EAD measures public opinion on political executives (mainly presidents) by combining multiple series and survey marginals into single country-level series. It uses Stimson’s (Reference Stimson1991) dyads-ratio algorithm for combining individual time series. This method should tackle some common problems related to aggregate-level public opinion data such as the usage of different questions and irregular survey cycles (Stimson, Reference Stimson2018). For example, some surveys rely on questions referring to general trust in the president, such as ‘Do you trust the president?’, while others use more direct questions about the performance of the incumbent president, such as ‘What grade would you give to the president in succeeding in their duties?’ According to Carlin et al. (Reference Carlin, Martínez-Gallardo and Hartlyn2012b: 213) Stimson’s (Reference Stimson1991) dyads-ratio algorithm is ‘the most ambitious application of this approach to the study of cross-national executive approval’. The algorithm is not applied to Finland as the series consists of only one regularly implemented survey.
The main explanatory variable is the constitutional strength of the presidency, measured as the level of presidential powers. A number of scholars have measured presidential powers with different techniques, yet there is no common understanding of which describes presidential powers the best. Shugart and Carey (Reference Shugart and Carey1992) developed one of the most used power scores, dividing presidential powers into legislative and non-legislative powers, an approach followed by many scholars while others have developed their own scores (e.g., Metcalf, Reference Metcalf2000; Siaroff, Reference Siaroff2003; Doyle and Elgie, Reference Doyle and Elgie.2016). In the context of semi-presidential studies, where presidential powers have been one of the key themes, Åberg and Sedelius (Reference Åberg and Sedelius.2020) identify altogether 37 studies of executive and presidential powers.
Following several scholars (e.g., Van Cranenburgh, Reference Van Cranenburgh2008; Elgie, Reference Elgie2011; Kim, Reference Kim2015), I rely on scores by Siaroff (Reference Siaroff2003), that include nine different sources of power (see Table 2). The benefit of using this scoring is that it understands presidential powers as a wider concept than just the constitution (Morgenstern et al., Reference Morgenstern, Perez and Peterson.2020). However, some of these categorizations are quite strict since, for example, Finland scores 0 in terms of ‘central role in foreign policy’ although the Constitution of Finland states that ‘The foreign policy of Finland is directed by the President of the Republic in co-operation with the Government’. At the same time, the total scores for each country seem to describe the institutional differences quite well, and, as expected, the scores are higher in presidential countries than in semi-presidential countries (with the exception of France and Peru) (see Table 2). In comparison, the presidential power scores by Doyle and Elgie (Reference Doyle and Elgie.2016) rate the Czech presidency as strong as the USA presidency, and France as the weakest semi-presidential country right after Finland, although many scholars would agree that despite the letter of constitution in France, the French president is politically very dominant in the national political system (e.g., Shugart, Reference Shugart2005) and therefore differs from many other presidencies within the regime type. Doyle and Elgie (Reference Doyle and Elgie.2016) combine 28 existing measures, which might lead to spurious results because of the different methods used by original authors, underlining the difficulty to measure power in general.
Presidential power scores by Siaroff
Categories by Siaroff (Reference Siaroff2003): PE = popularly elected; CE = concurrent election of president and legislature; AP = discretionary appointment powers; CM = chairing of cabinet meetings; VT = right of veto; EDP = long-term emergency and/or decree powers; FP = central role in foreign policy; GF = central role in government formation; DL = ability to dissolve the legislature. In the original scoring by Siaroff, the Czech Republic scored 0 for PE, yet as the dataset includes a time period for the Czech Republic from 2013 to 2020 it is therefore coded as 1 as direct elections for the president were introduced in 2012. Full description of the scores can be found in Siaroff (Reference Siaroff2003).
Siaroff himself has argued that ‘there is really no such thing as a semi-presidential system when viewed through the prism of presidential powers’ (Siaroff, Reference Siaroff2003: 307). This criticism is related to the concept of semi-presidentialism in general and its wide coverage of countries with different levels of presidential powers. Without going further into this debate, such a wide definition of semi-presidentialism certainly brings some challenges to a comparative study like this, which is why the level of presidential powers should be treated as the explanatory variable instead of the regime type. Quite interestingly, the original definition of semi-presidentialism by Duverger (Reference Duverger1980) states that the president should possess ‘quite considerable powers’ for the country to be considered as a semi-presidential system, which to some extent contradicts the argument about weak and popular presidents in semi-presidential regimes. Here the basic premise is that presidents in semi-presidential regimes are not ceremonial but usually weaker than presidents in presidential regimes, and that, for example, France and Peru can be treated as outliers differing from the general picture.
Other important independent variables are level of unemployment, GDP per capita growth, honeymoon period, and the end of the presidential term. The economic determinants are from the World Bank’s World Development Indicators and they enter the dataset as continuous variables. Increasing unemployment rates reflect weakening economic situations while increasing GDP reflects an improvement to it. Honeymoon period is measured as the first three months in the office after elections (1 = first quarter (first three months) or two first quarters if elections occur in the middle of a quarter; 0 = rest of the months), and the period at the end of the term is measured as the last three months in the office (1 = last quarter in the office or last two quarters if elections are in the middle of a quarter; 0 = rest of the months). Information on the presidential terms is collected from WorldStatesmen.org.
Countries enter the models as dummy variables to control for the effect of country-level differences. In addition, to control for the possibility that the general level of presidential popularity depends on the overall societal stability of the country, I include the level of corruption as a control variable. For example, in younger and less developed democracies such as in many European post-communist or Latin American countries, the citizens may express general distrust toward political institutions (e.g., Závecz, Reference Závecz, Zmerli and van der Meer2017), which may impact the general level of public opinion on presidents as well. To measure this ‘instability’, corruption should work as a proxy for such situations. The data is from the Corruption Perceptions Index by Transparency International and it is reverse-coded as, in the original data, lower values refer to higher levels of corruption and vice versa.Footnote 6 Other control variables such as the level of democracy, nonpartizan background of the president, and a dummy for the regime type (presidentialism/semi-presidentialism) were also considered yet only three presidents in the dataset had a nonpartizan ‘status’ (Worldstatesmen.org), while multicollinearity occurred between democracy and corruption and between the regime type and presidential powers.
Finally, several interaction terms are created to test the hypotheses further. The assumption is that the impact of economic and electoral cycle variables is conditional upon presidential powers. The interactions are presidential powers × GDP per capita growth, presidential powers × unemployment, presidential powers × honeymoon effect, and presidential powers × end of the term effect, and they enter the models in two stages.
Empirical analysis
The empirical section starts with descriptive statistics and figures of presidential popularity between countries and according to the level of presidential powers. Country-level figures of the popularity ratings and descriptive statistics of all the variables are in the online Appendix A. On average, less than half of respondents (around 45 percent) have a positive opinion of the president. The popularity ratings are slightly more positive in semi-presidential regimes than in presidential regimes. This relatively small difference between the regime types may possibly be explained by the variation in the level of presidential powers within semi-presidentialism. For example, the level of presidential popularity is extremely high and stable in Finland (one of the weakest presidencies within semi-presidentialism), and very low in Peru and France (the strongest ones). In terms of variation within individual countries (standard deviation and range), the scores seem a bit more unstable in presidential regimes than in semi-presidential regimes. This might, however, be related to the number of individual presidents and the fact that presidents enter and leave office more frequently in presidential regimes (see Table 1 in previous section).
Moving to the effect of presidential powers, Figure 1 presents presidential popularity ratings and presidential power categories (Siaroff, Reference Siaroff2003) as a group comparison. The graph clearly shows that presidents with weaker powers tend to be more popular than presidents with stronger powers. Only the strongest presidents in category 8, including Argentina, Brazil, and Chile, stand out with higher levels of popularity than in the previous category, yet on average they are still less popular than their weaker counterparts in semi-presidential regimes (categories 2 and 3), and receive sometimes extremely low popularity ratings. Results of the ANOVA test (reported under Figure 1) show that there is a significant difference in the mean values between the groups.
Presidential popularity and presidential power scores, mean comparison. F-values by ANOVA = 164.2***(P < 0.001). Siaroff’s (2003) power scores: 2 = Czech Republic, Finland; 3 = Bulgaria, Portugal, Poland; 6 = South Korea; 7 = France, Peru, Costa Rica, Panama, USA, Uruguay; 8 = Argentina, Brazil, Chile.
In contrast, a comparison between the regime types reveals that the differences between the popularity ratings are not as clear when not accounting for the level of presidential powers (Figure 2). In fact, although presidents in semi-presidential regimes receive the highest popularity ratings and presidents in presidential regimes receive the lowest, the difference in the mean values between the regime types is not statistically significant (see t-test). Additionally, when looking at the trend of the popularity ratings in a time series figure (Appendix A4), the general level is not significantly higher in semi-presidential regimes than in presidential regimes, being even lower for several years from 2008 to 2015, which is interesting in light of the financial crisis over the same time period. The gap between the regime types is, however, larger and the effect of the time period between 2008 and 2015 is smaller when France and Peru, representing strong presidencies among semi-presidential regimes, are removed from the comparison. Similarly, removing Finland brings the series closer to each other. This further supports the assumption that the constitutional strength of the presidency should be a more important factor in terms of public opinion on presidents than just the classification of the regime types.
Presidential popularity and regime types, mean comparison. t-value by two-sample t-test = −0.846 (P > 0.1).
Table 3 reports the results of the regression coefficients with robust standard errors. There are altogether three models, of which Models 2 and 3 show the conditional effects of the economic variables and electoral cycle. First, Model 1 shows a strong and positive impact of presidential powers, GDP per capita growth, and the honeymoon period on presidential popularity ratings, while unemployment rate shows a negative impact. When the interaction terms between the electoral cycle and presidential powers are entered in Model 2, the individual effect of honeymoon disappears but the interaction term turns out to be statistically significant, indicating that the honeymoon effect may be dependent on the level of presidential powers. The effect of the end of the term does not, however, turn out to be statistically significant individually nor conditionally. Similarly, when the second set of interactions regarding the economic variables enter the equation in Model 3, the individual effects of both unemployment and GDP disappear, while both interaction terms turn out to be statistically significant. Again, this indicates that in terms of public opinion on presidents, the negative impact of unemployment and positive impact of GDP are dependent on the level of presidential powers. Level of corruption (control variable) is weak and statistically insignificant in all three models.
Determinants of presidential popularity, fixed effects regression models
*P < 0.05.
**P < 0.01.
***P < 0.001.
All variables are standardised between 0 and 1. Robust standard errors are in parentheses.
The results thus indicate that a higher level of presidential powers tends to decrease the level of presidential popularity. This is supported by the consistent individual effects of the variable (confirming H1) and the conditioning effects when interacting with the economic variables and the honeymoon period (confirming H2 and H3). Furthermore, without the interactions, the honeymoon period and the economic determinants show clear effects on presidential popularity, as suggested by earlier literature, yet when the level of presidential powers is considered, the picture changes, and the effects of these variables become conditional on higher levels of presidential powers.
For robustness, the same three models were performed with several adjustments. The results of these additional robustness checks are in Appendices B1, B2, and B3. First, Siaroff’s (Reference Siaroff2003) presidential power scores were replaced by the scores of Doyle and Elgie (Reference Doyle and Elgie.2016) as the main independent variable. Despite the difficulties related to these power scores, the models produced very similar results, indicating that the constitutional strength of the presidency is clearly reflected in the presidents’ popularity ratings. With the measures by Doyle and Elgie (Reference Doyle and Elgie.2016), the connection between popularity and powers was even stronger, although the conditional effect of GDP per capita growth turned out to be statistically significant instead of unemployment when considering the level of presidential powers.
Second, the dependent variable was replaced with lagged values to avoid autocorrelation problems, as suggested by several studies (Berlemann and Enkelmann, Reference Berlemann and Enkelmann.2014: 48). Both two and four lags were tested (the former is reported in the Appendix), and the results showed again a largely similar outcome, especially in terms of presidential powers and the economic variables. However, the coefficients of the electoral cycle showed mixed results, which is not surprising as both the honeymoon period and the end-of-the-term boost happen in short cycles.Footnote 7
Third, results of an Ordinary Least Squares (OLS) regression model without country dummies are reported since many studies of the determinants of public opinion on presidents rely on this basic method (Berlemann and Enkelmann, Reference Berlemann and Enkelmann.2014: 44). The outcome was of course impacted by the removal of the country-effects, but the main results remained the same: presidential powers showed strong explanatory power in terms of individual and conditional effects. As a difference, corruption showed a strong negative effect on presidential popularity, which may logically result when the within-country variation is not considered.
Discussion
Public opinion plays a crucial role in democracies. Opinion polls are followed closely by political parties and the media, and changes in them may impact election results and the political atmosphere between elections. Whether the presidents are able to use their popularity for policy purposes or not, increasing or decreasing popularity ratings may impact the behavior of the presidents and other political actors. This makes even weaker presidents’ popularity functions worth studying.
This article has focused on patterns of presidential popularity between weaker and stronger presidents across different constitutional settings, addressing an evident research gap. Furthermore, while the majority of presidential popularity studies have focused on presidents in presidential regimes, this study examined public opinion within and between presidential and semi-presidential regimes with presidential powers as the main possible explanation. The most important implication of this study is that the level of presidential powers has an impact on how presidents are perceived by the people. Moreover, the results show that the general level of presidential popularity tends to be higher among presidents with weaker constitutional prerogatives, confirming hypothesis H1. Given this, presidential powers should be included as one of the main explanatory variables when studying the aggregate-level determinants of public opinion on presidents across political systems.
The article not only offered empirical evidence of the connection between presidential powers and presidential popularity but also revealed that constitutionally stronger and weaker presidents are judged differently when it comes to the effect of changes in the national economy or the electoral cycle, confirming hypotheses H2 and H3. Furthermore, the impact of these factors was conditioned by the level of presidential powers. Whereas presidents with stronger constitutional powers suffer from increasing unemployment rates and benefit from higher GDP per capita growth, presidents with weaker constitutional powers seem to be somewhat above these trends in the national economy. A possible explanation is that changes in the economy are reflected more in the popularity ratings of the prime ministers and governments because they are responsible for these policy areas. This finding is linked to the responsibility theory suggesting that short-term changes in the political environment, such as the state of the economy, affect presidential popularity because people hold incumbent officeholders responsible for these matters (e.g., Nannestad and Paldam, Reference Nannestad and Paldam.1994). Similarly, not being at the center of attention regarding domestic policy issues, weaker presidents do not seem to suffer from a decline after the beginning of their term as much as their stronger counterparts, as reflected in the results of the honeymoon effect. In other words, the cyclical patterns of presidential popularity found in the context of stronger presidencies do not seem to have as strong an explanatory power in the context of constitutionally weaker presidents.
Countries selected for the analysis were either presidential or semi-presidential. The presidential countries were quite homogeneous in terms of the strength of the presidential institution, while the semi-presidential ones covered a full variety of powers in different political systems from long-developed stable democracies to relatively new and politically more unstable systems. Some might wonder whether the regime classifications matter in this case at all. However, as already stated, it is not the regime type itself that determines the level of presidential popularity, but rather the division of powers in such regimes that enables the president to situate and stand out from other political institutions in a way that benefits the president. At the same time, parliamentary regimes with indirectly elected and fully ceremonial presidencies were not included in the analysis. These systems could work as an interesting reference and widen the spectrum of presidential powers even further. For example, if the presidents’ role ‘above party politics’ is reflected in their popularity ratings, how are purely ceremonial presidents or even monarchs in parliamentary monarchies being judged?
One challenge relates to the shortcomings of aggregate-level data and the lack of cross-national longitudinal panel surveys of presidential popularity. However, including a large number of countries, a long enough time period, targeted contextual variables, and several robustness checks in the empirical analysis ensured the reliability of the results. At the same time, although the general picture of weak but popular presidents seems to apply here, the link between presidential powers and public opinion should be examined further with individual-level data as well, not least from the perspective of citizens’ beliefs and perceptions of the presidential institution in general.
Finally, this study addressed systemic-level patterns regarding the level of presidential popularity in the context of the level of presidential powers without going into details about the various types of the powers. With a more comprehensive dataset, the impact of the powers of the presidents in specific policy areas or processes could be examined across systems to trace which of the powers make presidents less popular. Patterns of presidential popularity should also be examined more systematically in the context of presidential activism – the actual behavior of the presidents and how they utilize their constitutional powers.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1755773923000280
Acknowledgments
I would like to thank the anonymous reviewers for their insightful comments and suggestions, the discussants and participants at the 2022 European Consortium for Political Research (ECPR) General Conference in Innsbruck where a previous version of this article was presented, and my colleagues at Tampere University for their feedback and support.
Funding statement
The work was funded by the Academy of Finland, grant number: 333013.
Competing interests
The author declares none.
Open access
