Opinion polls, political connections, and firms

Universidade do Minho Escola de Economia e Gestão Bruno Emanuel Pires Fernandes Opinion Polls, Political Connections, and Firms outubro de 2024 Opinion Polls, Political Connections, and Firms Bruno Fernandes UMinho | 2024 Universidade do Minho Escola de Economia e Gestão outubro de 2024 Bruno Emanuel Pires Fernandes Opinion Polls, Political Connections, and Firms Tese de Doutoramento Doutoramento em Economia Trabalho efetuado sob a orientação do(a) Professora Doutora Linda Gonçalves Veiga Professor Doutor Luís Aguiar-Conraria ii DIREITOS DE AUTOR E CONDIÇÕES DE UTILIZAÇÃO DO TRABALHO POR TERCEIROS Este é um trabalho académico que pode ser utilizado por terceiros desde que respeitadas as regras e boas práticas internacionalmente aceites, no que concerne aos direitos de autor e direitos conexos. Assim, o presente trabalho pode ser utilizado nos termos previstos na licença abaixo indicada. Caso o utilizador necessite de permissão para poder fazer um uso do trabalho em condições não previstas no licenciamento indicado, deverá contactar o autor, através do RepositóriUM da Universidade do Minho. Licença concedida aos utilizadores deste trabalho Atribuição-NãoComercial-SemDerivações CC BY-NC-ND https://creativecommons.org/licenses/by-nc-nd/4.0/ iii Acknowledgments I gratefully acknowledge the financial support from the Portuguese Foundation for Science and Technology (FCT) through the SFRH/BD/138513/2018 fellowship and the support of my host institution, NIPE. I am eternally grateful to my supervisors. Their knowledge, guidance, and advice were essential to my growth as a young researcher. I am indebted to Linda Veiga, who has accompanied me since my master’s thesis as a supervisor. I owe it to her if there is any quality in me as a researcher. I am also indebted to Luís Aguiar-Conraria. His straightforwardness and his practical sense helped me to make progress in my research and to overcome obstacles. I would also like to express my sincere gratitude to all those who have accompanied me on this journey. Professors, colleagues, and friends have been essential to my success and personal and professional growth. I must especially mention Hélder, João, and Rui, who have been with me since the beginning of my undergraduate studies. We have shared great memories and countless study hours, which I consider priceless. I am deeply thankful to my family, who have always supported me and shown me the right direction. A special thank you to my grandparents, who have unfortunately already passed away but who I believe would be proud of my journey. I want to thank my parents and sister for their unconditional love and support. I want to express my gratitude to Margarida for everything she has done for me, for her unconditional support and motivation, and for sharing the good and bad times with me. Without her, it would not have been easy to reach the end. iv STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Code of Ethical Conduct of the University of Minho. v Sondagens, Conexões Políticas e Empresas - Resumo Um dos temas mais debatidos em Economia Política é o impacto da economia na popularidade dos líderes políticos. É amplamente aceite que esta tem impacto na popularidade, no entanto, a investigação tem produzido resultados instáveis e, por vezes, contraditórios. Os três primeiros ensaios desta dissertação procuram dar foco a potenciais causas deste problema. O primeiro ensaio estuda a diferença entre a economia “objetiva” e “mediada”. O impacto da evolução verificada da economia na aprovação dos políticos pode ser menor do que o impacto da economia retratada nos media . Com uma base de dados para Portugal, para o período de 2001 a 2018, analisamos se a aprovação melhora quando utilizamos a informação económica que é disponibilizada ao público. Concluímos que esta é prevista com maior rigor quando usando o crescimento económico projetado para o ano em curso. Num sistema semipresidencial, a aprovação quer dos primeiros-ministros, quer dos presidentes, pode ser influenciada pela economia. Usando uma base de dados para Portugal entre 1986 e 2018, o segundo ensaio estuda a possibilidade de os efeitos económicos serem assimétricos, ou seja, períodos de crise a exercerem efeitos mais significativos, e se tais efeitos são mitigados em períodos de coabitação. Os resultados sugerem que a coabitação e a falta de incentivos à reeleição para os presidentes, contribuem para obscurecer a responsabilização dos políticos pelos resultados económicos. Adotando uma abordagem diferente, o terceiro ensaio aprimora o uso das ferramentas wavelet na ciência política, alargando o foco para a análise multivariada e introduzindo estimações de equações usando wavelets. O ensaio explora o impacto de três variáveis económicas na popularidade do último presidente dos EUA a cumprir dois mandatos, Barack Obama. Os resultados mostram que a sua aprovação está estruturalmente ligada às taxas de desemprego e de inflação, com diferentes padrões de sincronização, e relacionada de forma transitória com o sentimento do consumidor. Finalmente, o quarto ensaio aborda a relação entre política e negócios, um tema multidisciplinar. Investiga as características e o impacto das conexões políticas nas empresas cotadas na bolsa portuguesa entre 2002 e 2022, estimando a probabilidade de ter conexões políticas e o impacto nas variáveis financeiras. Os resultados do ensaio sugerem que as conexões políticas são mais comuns em empresas de maior dimensão, especialmente em indústrias reguladas e localizadas na área de Lisboa. No entanto, o impacto financeiro dessas conexões parece ser limitado. Palavras-Chave: Análise com Wavelets; Coabitação Política; Conexões Políticas; Funções Popularidade; Variáveis Económicas vi Opinion Polls, Political Connections, and Firms - Abstract One of the most studied and debated topics in the field of Political Economy is the impact of the economy on how voters evaluate their political leaders. It is widely accepted that the economy influences political popularity, yet research has often produced unstable and sometimes contradictory results. The first three essays of this dissertation aim to shed light on the potential causes of this problem. The first essay studies the difference between the “objective” and the “mediated” economy. Economic developments on the ground may less influence voters’ approval of incumbents than the economy portrayed by the media. Using evidence from Portugal from 2001 to 2018, we examine whether the ability to predict approval is improved when we employ economic information made available to the public. We conclude that approval is best predicted by forecasted growth for the current year. In a semi-presidential system, the approval of prime ministers and presidents can be affected by economic performance. The second essay uses approval data and GDP growth figures from 1986 until 2018 in Portugal. We examine the possibility that economic effects are asymmetric , with bad economic times exerting stronger effects, and whether such effects are suppressed in periods of cohabitation. Results suggest that cohabitation combined with a lack of reelection incentives for presidents contribute to obscure accountability for economic outcomes, diminishing the role of economic performance. Taking a different approach, the third essay enhances the use of the wavelet toolset for political science by extending the analytical focus from bivariate to multivariate and introduces regression estimation using wavelets. It explores the impact of three economic variables on the popularity of the most recent two-term US President, Barack Obama. The results show that Obama’s approval rating was structurally linked to unemployment and inflation rates, with different synchronization patterns, and transiently related to consumer sentiment. Finally, the fourth essay addresses the relationship between politics and business, a multidisciplinary topic. It investigates the characteristics and impact of political connections on Portuguese publicly listed firms between 2002 and 2022 and estimates the probability of political connections and their effect on firms’ financial outcomes. The results suggest that political connections are more common in larger firms, especially in regulated industries and those located in the Lisbon area. However, the overall financial impact of these connections appears to be limited. Keywords: Economic Variables; Political Cohabitation; Political Connections; Popularity Functions; Wavelet Analysis. vii Table of Contents 1. Introduction ................................................................................................................................ 1 2. Economic Forecasts and Executive Approval ........................................................................ 5 2.1. Introduction ........................................................................................................................ 5 2.2. The “Real” and the “Mediated” Economy ...................................................................... 6 2.3. Data and Methods .............................................................................................................. 9 2.4. Results ............................................................................................................................... 14 2.4.1. Vintage vs. First Values ............................................................................................ 14 2.4.2. The Forecasted Economy......................................................................................... 16 2.5. Conclusion ......................................................................................................................... 18 3. The Economy and Executive Approval in a Semi-Presidential Regime: the Case of Portugal ......................................................................................................................................... 20 3.1. Introduction ...................................................................................................................... 20 3.2. Public approval and the economy in Portugal .............................................................. 22 3.2.1. Economic Contexts and Asymmetry ....................................................................... 22 3.2.2. Portuguese Semi-Presidentialism and its Implications ........................................ 23 3.3. Data, Variables, and Estimation Strategy ..................................................................... 26 3.3.1. Data and Variables ................................................................................................... 26 3.3.2. Estimation Strategy .................................................................................................. 30 3.4. Results ............................................................................................................................... 31 3.5. Conclusion ......................................................................................................................... 36 4. US Presidential Approval Rating: Time-Frequency Approach ............................................ 39 4.1. Introduction ...................................................................................................................... 39 4.2. Wavelet analysis and the study of popularity functions .............................................. 40 4.2.1. The challenge of instability ...................................................................................... 40 4.2.2. The (partial) wavelet coherency, gain, and phase-difference ............................ 41 4.3. An example ........................................................................................................................ 43 4.3.1. The Data Generating Process ................................................................................. 43 4.3.2. Reading the results ................................................................................................... 44 4.4. Data .................................................................................................................................... 46 3 In times of economic turbulence, the impact of the economy on popularity is perceived to be significant. Thus, in Chapter 4, I studied the presidency of Barack Obama, characterized by a slow economic recovery after the subprime mortgage crisis in 2007. To do so, I used two important economic variables, the unemployment rate and the inflation rate, as proxies for the state of the economy, and an additional variable able to catch voter’s perceptions, which is the consumer sentiment, sourced by the Michigan University. Comparing the results of OLS estimations and the results of wavelet analysis, running estimates that are time-varying can shed some light on the reasons why the relationship between economy and political popularity has not been consistent. Politicians care about their popularity because they desire to win elections. Maximizing the probability of being elected is a priority for the incumbent. The literature on political business cycles is abundant, stating that politicians make an effort to accelerate economic performance by reducing unemployment or fostering the economy or by increasing visible expenditures when the elections draw near. One mechanism by which the incumbent may use to cause this political cycle is through political connections. Politically influenced firms can benefit incumbents around the election time by fostering employment (Bertrand et al. , 2018) or contributing directly or indirectly with funds to the campaign (Claessens, 2008; Sukhtankar, 2012). However, politicians also have interests other than electoral or political. They have their own private interests. Therefore, it is not difficult to find politicians as members of firms’ boards or other relationships of different natures (e.g., lawyers, consultants, or auditors). Firms also benefit from establishing these connections with politicians in a quid pro quo relationship. Otherwise, they would not exist (Faccio and Hsu, 2017). They can be benign, with connected firms reaping the benefits of having reputed people who know how the government administration works, or malign. Some firms are interested in establishing connections to obtain advantages against their competitors, which could be associated with rent-seeking activities or corruption. A survey conducted in 2024 by the European Union (EU) to measure “[…] the level of corruption perceived and experienced by businesses […]” indicates that 40% agree with the idea that in their country, the only way to succeed in business is to have political connections (European Commission, 2024a). The percentage becomes more pronounced with the Portuguese businesspeople, with 65% agreeing with the statement (the highest value among the EU countries). In addition, the majority of the EU businesspeople (79%) agree that links between businesses and politics lead to corruption. The percentage increases to 82% when we restrict the analysis to Portuguese businesspeople. Another survey conducted in 2024 by the EU to evaluate “[…] the level of corruption perceived and experienced by European citizens […]” 4 highlights that 51% agree that political connections are key to succeed in business, and the percentage increases to 55% among Portuguese citizens (European Commission, 2024b). As in the survey for businesses, 75% of European citizens agree that links between business and politics are associated with corrupt behavior. Within the Portuguese citizens, the percentage increases to 81%. The perceived link between political connections and corruption undermines voters’ confidence in politicians and influences their evaluation of how good or bad politicians are handling their jobs (Anderson and Tverdova, 2003). However, this perception does not prevent politicians from being elected, even if they have been accused or condemned for corruption (de Sousa and Moriconi, 2013). Literature has studied political connections and the complex network between politics and businesses. Many of these networks are associated with rent-seeking behavior and corruption, which have a pervasive impact on economic outcomes (Schoenherr, 2019; Khwaja and Mian, 2005). Often, this relationship involves a quid pro quo dynamic characterized by reciprocal exchanges and benefits: increasing political influence for politicians or extracting rents, easing regulations, or creating barriers to entry for incumbent firms. The presence of politicians and former politicians on company boards is a global phenomenon, not exclusive to developing countries. In Portugal, it is common to find former politicians on company boards. Chapter 5 of this dissertation delves into the existence and characteristics of political connections in Portugal. Studying these connections in Portugal is important because it allows researchers to analyze their impact on the economy. This chapter examines political connections in publicly traded Portuguese companies from 2002-2022. Firm size and regulated sectors are associated with political ties, particularly those headquartered in Lisbon. Additionally, right-wing politically connected members are common in larger firms, whereas left-wing politically connected members are more noticeable in firms whose headquarters are located in the Lisbon area. However, the analysis shows no significant impact of political connections on financial performance. Finally, Chapter 6 summarizes the main findings from the dissertation, discusses limitations, and offers suggestions for future research. 5 2. Economic Forecasts and Executive Approval 1 2.1. Introduction One of the most established ideas in the study of public opinion and elections is that voters care about the economy when evaluating incumbents (Lewis-Beck and Stegmeier, 2000 and 2013). Yet, objective economic performance indicators, such as GDP growth, unemployment, or inflation, do not seem to be very strongly related to political support as expressed in executive approval or voting for the incumbent. Although “the (real) economy matters,” “its effects are weak and contain little explanatory power” (Van de Eijk et al. , 2007: 180). “[T]he economy is only a small determinant of electoral outcomes (…) and captures a tiny proportion of the overall variation in support for the lead party” (Kayser, 2014: 117). Some even argue that, at least for the wealthier and more established democracies, economic growth may have no effect on the electoral fate of incumbents (Brender and Drazen, 2007). How can we reconcile the seemingly obvious notion that the economy is essential for voters with the weak explanatory power of objective economic indicators? In a 2015 study, Kayser and Leininger (2015) use an imaginative research design to address this puzzle. Taking the case of the United States, they estimate different election forecasting models employing vintage data on GDP growth (the revised historical estimates of economic performance, reflecting the most accurate measurements of economic activity available) and real-time data on GDP growth (the first estimates of economic performance released immediately after the period to which they are referring and circulated in the media). They conclude that the data that most accurately reflects the state of the economy (the vintage data) ends up being a worse predictor of political support than the less accurate initial economic estimates conveyed by the media (the real-time data). In other words, the best measure 1 A version of this Chapter was published as: Aguiar-Conraria, L., Fernandes, B., & Magalhães, P. C. (2024). Economic forecasts and executive approval. Journal of Elections, Public Opinion and Parties, 34(4), 643-655. 6 of “real” economic growth and the mediated information about economic growth available to voters do not necessarily match, and the latter is more politically consequential than the former. This essay draws inspiration from Kayser and Leininger’s approach but to test yet an alternative answer as to why objective indicators about the recent performance of the economy may not be very good predictors of political support: the available information people are more likely to use to evaluate incumbents is not about recent economic performance, but instead about expected future performance. In other words, voters rely upon information of even more uncertain accuracy than real-time data to evaluate incumbents, and they do so prospectively rather than retrospectively. We test this hypothesis for the case of Portugal by looking at different series of economic data as predictors of prime ministerial approval. First, we consider the most recently available Statistics Portugal (INE) historical time-series data on the year-on-year GDP growth rate for each quarter, which have been revised for accuracy. However, we also consider both the preliminary — less accurate — estimates of the same GDP growth announced by Statistics Portugal in their quarterly publications since 2000 and the OECD’s forecasts of GDP growth obtained from the organizations’ Economic Outlook issued in the second and fourth quarters of each year . We show that the information made available over time about the forecasted growth for the current year is the best predictor of prime ministerial approval. In other words, like Kayser and Leininger (2015), we find indirect evidence that voters respond less to the “objective economy” than they do to the information that happens to be available about it. However, in the case of Portugal, the most consequential available economic information seems to be not about the past but rather about the near future. The chapter is structured as follows. Chapter 2.2 discusses the literature on popularity functions and the media's impact on people’s expectations and perceptions. Chapters 2.3 and 2.4 describe the data and methodology used and the econometric results. Chapter 2.5 concludes. 2.2. The “Real” and the “Mediated” Economy A long tradition in political science, initiated by Mueller (1970) and Goodhart and Bhansali (1970), has shown that voters’ support for the executive is related to economic performance. GDP growth, inflation, and levels of/changes in unemployment rates have all been shown, at one time or another, to account for part of the variance in the support awarded by the public to the executive, as captured by surveybased indicators or by the vote in the incumbent party or president. 7 However, since those founding works, research has increasingly moved away from a conception of “a public that almost mechanistically responds to economic conditions” (Gronke and Newman, 2003: 506). The central aspect of this movement is the notion that what should ultimately be consequential for political support is voters’ subjective perceptions of economic performance. Those perceptions, although themselves shaped by economic conditions (Becher and Donnelly, 2013), are also affected by a myriad of other factors besides the “real” performance of the economy. These include voters’ political predispositions (Wlezien et al., 1997; Evans and Andersen, 2006; Chzhen et al., 2014), news coverage, and people’s exposure to it (Goidel and Langley, 1995; Sanders and Gavin, 2004; De Boef and Kellstedt, 2004; Boydstun et al., 2018), and combinations of the above (Duch et al. , 2000). This basic idea at once clarifies and complexifies the relationship between the economy and political support. On the one hand, it suggests that the substantively small correlations between economic aggregates and political support that have emerged in the literature should not be interpreted as signaling that voters do not care about the economy. Instead, they suggest that the study of the economic drivers of political support was never primarily about the “real” or the “measured” economy but rather about the economic information made available to different voters by the media — the “mediated economy” — and how those voters deal with that information to form subjective perceptions (Stevenson and Duch, 2013). On the other hand, relying on subjective economic perceptions as predictors of political support raises the risk of overestimating how much economic perceptions really matter for voters when evaluating incumbents, particularly if we consider how much those perceptions may be driven by political support itself and how difficult it is to address the related measurement and endogeneity problems in observational studies (Anderson, 2007, and many others). This dilemma remains, until today, one of the most contentious in the sub-field (Hellwig and Marinova, 2017). However, the importance of the “mediated economy” can be shown, at least indirectly, in ways that circumvent such threats to inference. Kayser and Leininger (2015) examine the predictive power of different forecasting models for U.S. presidential elections that employ economic variables. As economic input, they use two types of indicators. The first, Vintage , corresponds to GDP or GNP growth estimates as revised by the U.S. Bureau of Economic Analysis. The second, Real-Time , is the early estimates of GDP or GNP growth for each quarter published one to two months after the end of that quarter. In other words, while real-time data capture the economic information available to voters as each election approaches, as conveyed by the media, vintage data correspond to enhanced measurements that more accurately capture economic activity and performance. They find that vintage data introduces greater forecasting errors than real-time data and that the error increases as the number of revisions increases. 8 In other words, the more accurate the economic data, and the more different they become from the initial estimates, the worse predictors of political support they become. Indirectly, this suggests that voters’ response to economic performance is a response not so much to the “objective” experience of how the economy has performed but rather to information about economic performance as conveyed through the media. However, taking the mediated economy seriously forces us to consider that it is not composed exclusively of information about recent economic performance. It also contains information about the future, including forecasts of economic performance from professional economists, the government, central banks, or international organizations. Several studies have shown that, at least in countries like the US, the UK, or Canada, the positivity and negativity of economic news seems to mirror more closely leading rather than current or past economic indicators. In other words, the tone of economic news is more a reflection of future economic developments than past ones (Soroka et al. , 2015; Wlezien et al. , 2017). Correspondingly, media messages about the economy seem to shape not only people’s subjective perceptions about how the economy has been doing but also their subjective expectations about how the economy is likely to perform in the future (Soroka et al. , 2015). Some even argue that media messages’ primary role is to shape expectations about a more uncertain future rather than affecting evaluations of an already experienced past (Boomgaarden et al., 2011; Damstra and Boukes, 2021; Boukes et al. , 2019). And there is, of course, considerable evidence that past economic performance or people’s perception of it are not necessarily the best predictors of political support. Instead, voters may behave like “bankers,” judging governments based on expectations regarding the future economy (MacKuen et al. , 1992; Erikson et al. , 2000), at least for some voters (Alt et al. , 2016; Lacy and Christenson, 2017; Acevedo et al. , 2017), in some contexts (Singer and Carlin, 2013), and in some political systems (Cohen, 2004). That the “mediated economy” also contains information about the future and that voters care about such information is a possibility not contemplated by Kayser and Leininger’s (2015) comparison of “vintage” and “real-time” economic indicators in predicting political support. In their approach, voters who have “experienced the past” or are informed about it are assumed to derive political support from that experience or information. However, if economic information is also shaped by economic forecasts and/or voters are prospective instead of retrospective, information about the future should also matter to explain political support. This is exactly what we propose to examine in this study: besides examining the 9 role of different measures about the recent performance of the economy, we also examine the role of forecasts of that performance. 2.3. Data and Methods Our dependent variable is a measure of prime ministerial approval in Portugal available at the Executive Approval Project (Carlin et al., 2019). We use quarterly data from 2001 to 2018, resulting from polls conducted by Portuguese pollsters. We use a smoothed version of the variable Approval (% of positive ratings) available in the Executive Approval Database 2.0 for Portuguese Prime Ministers. Figure 1 plots the evolution of prime ministerial approval from the first quarter of 2001 until the third quarter of 2018, covering eight governments and six prime ministers. To measure economic performance, we consider several different variables to capture the growth rate of Portuguese’s real GDP (Table 1): GDP Vintage Growth , GDP First Values Growth, GDP Real-Time Figure 1: Quarterly data on approval rating for Portuguese Prime Ministers from the first quarter of 2001 to the third quarter of 2018. Note: Each dashed line corresponds to a government’s inauguration. Source: Carlin et al. (2019). 10 Growth, OECD Forecast Current, and OECD Forecast Next . 2 Because the differences are subtle, we will describe them in detail. GDP Vintage Growth represents the most accurate data on the evolution of real GDP, the revised values for each quarter. This information is conveyed by the Portuguese Statistics Institute (INE). We use the revised data published in the first quarter of 2021, representing the best information available regarding the evolution of real GDP throughout the entire 2001-2018 period. For the purpose of our econometric analyses, we take into consideration that voters need time to internalize economic information and thus assume a conventional short lag between the economic measures and approval (one quarter, t-1). Table 1: Economic variables employed. Variables Definition Source GDP Vintage Growth t-1 2021 revision of the year-on-year real GDP growth rate at the quarter before approval was measured. Statistics Portugal (INE) GDP First Values Growth t-1 First estimates made public of the year-onyear real GDP growth rate at the quarter before approval was measured. Statistics Portugal (INE) GDP Real-Time Growth t-1 The most recent publicly available quarterly year-on-year real GDP growth rate at the quarter before approval was measured. Statistics Portugal (INE) OECD Forecast Current t-1 The most recent yearly forecast of real GDP growth for the current year publicly available at the quarter before approval was measured. OECD Economic Outlook OECD Forecast Next t-1 The most recent forecast of real GDP growth for the following year publicly available at the quarter before approval was measured. OECD Economic Outlook GDP First Values Growth t-1 is constructed using the very first values of the year-on-year quarterly real GDP growth rate disclosed by INE in their reports issued since 2000. The difference between vintage or revised and first values of GDP growth rate is observed in Figure 2. At first glance, the preliminary values tend to be more conservative than revised values. Again, in the analysis, we match approval rates 2 Economic indicators such as unemployment or inflation rates have also been frequently included in (early) studies of Portuguese government approval functions (Veiga 1998; Veiga and Veiga, 2004a). However, while inflation remained remarkably stable in Portugal throughout the period under examination, unemployment rates, as Goulart and Veiga (2016) argue, were somewhat deceptive indicators of economic performance, particularly in periods when massive emigration (such as during the 2008-2013 economic and financial crisis, when it reached an historical peak) contributed to mask the true magnitude of the crisis as expressed in unemployment rates. 11 with the value of GDP First Values Growth for the preceding quarter. However, if we take the point of view of when information is made available to the public, we should consider the fact that the GDP growth announced to the public by Statistics Portugal in each quarter refers not to the quarter of publication but to previous ones. Therefore, we also include, for each quarter preceding the one when approval is gauged, the most recent publicly available value for the year-on-year GDP growth: it is GDP Real-Time Growtht-1 . Finally, we consider the role of economic forecasts. We extract them from OECD’s reports, i.e., “OECD Economic Outlook.” Those forecasts are conveyed twice per year, every second and fourth quarter. In these publications, the OECD discloses annual real GDP forecasts for each member country for the current and following years. Although several institutions produce growth forecasts, the OECD provides the most extended series. 3 Its projections are always issued using the same timing. They consistently project growth for the current and the next years and are a highly reputed source heavily covered by the Portuguese media. They are influential in both news content and opinion articles. From 2000 to 2018, Lusa , the sole Portuguese National News Agency and a primary source for other media 3 Other options would have been the projections of the IMF or the European Commission. However, while the former has more irregular publication dates in comparison with the OECD, the latter has changed publication schedules in both 2012 and 2018. Moreover, Portugal received aid from both institutions between 2011 and 2014, making them significant political entities, which could affect the perceived impartiality of their forecasts in comparison to the OECD. Figure 2: Quarterly data on Real GDP Growth Rate, percentage change from the year ago. Note: The full line represents the very first values of quarterly real GDP growth, and the dashed line plots revised values of quarterly real GDP growth. Source: INE. 12 outlets, produced 97 news articles with the phrase “economic forecasts” in the headline and nearly 2500 mentions of this term within the body of the text. The specific combination between the same expression and “OECD” occurs in 416 news items in Portuguese news sources available in Factiva in the same period. 4 Our logic remains the same: we focus on the information available to the public in the quarter before prime ministerial approval is gauged. Therefore, we start by modeling PM Approval at the first and second quarters of each year as a function of the most recent OECD’s growth forecast for that year (which was issued in the last quarter of the preceding year), while PM Approval at the third and fourth quarters of each year is modeled as a function of growth projection (for that year) issued in the second quarter: OECD Forecast Current t-1. Finally, we also consider the possibility that PM Approval may be affected by information about forecasts for the year after which approval is gauged. Correspondingly, this variable follows the same structure as the previous one, but now we consider the GDP growth forecast for the following year: OECD Forecast Next t-1. The plots of these forecasts can be observed in Figure 3. As suspected, forecasts for the current year tend to be more accurate than forecasts for the year beyond. 4 We are grateful to Tiago Dias, from Lusa , and to Nelson Santos, from ICS-ULisbon, for sharing these data. Figure 3: Quarterly data on annual OECD forecasts and annual real GDP final values. Note: The full line shows projections for the current year, and the dashed line plots the projections for the next year. Source: OECD. 19 Between 2001 and 2018, Portugal experienced three recessions. The third one, a sovereign debt crisis, was severe enough to trigger a request for international financial assistance. As a result, it is not unreasonable to suggest that the Portuguese electorate may have been particularly attentive to the economic projections presented by international organizations such as the OECD. To ensure the generalizability of our findings, further investigations should be conducted in various regions. A possibility for future research would be to conduct a panel study, taking advantage of the consistent production of the OECD Economic Outlook across several countries. Furthermore, a complete test of this argument would have to rely on measurements of individuallevel perceptions, the way they are shaped by economic forecasts and their media coverage, and, in turn, how they drive executive approval. Still, these results suggest the importance of economic forecasts in shaping public attitudes. They also encourage possible new approaches to modeling in election forecasts, an important field of political economy studies (Stegmaier and Norpoth, 2013). Typically, models that include economic variables resort to “vintage” data for estimation and analysis of fit and then plug “realtime” data to produce concrete forecasts. To the extent that the lessons drawn here about executive approval extend to electoral support for incumbents, economic forecasts may capture better citizens’ overall perceptions of the state of the economy and how they use those perceptions to evaluate political performance. 20 3. The Economy and Executive Approval in a Semi-Presidential Regime: the Case of Portugal 10 3.1. Introduction Over the last two decades, several studies have enhanced our understanding of the relationship between the economy and public support for the executive in Portugal (Veiga 1998; Veiga and Veiga, 2004a and 2007). The central finding is clear: the public approval of governments and prime ministers is affected by economic performance. This is not at all surprising. Left-right socio-economic policy issues have almost wholly dominated citizens' concerns throughout Portugal's democratic history (Tsatsanis et al. , 2014: 527). Despite a proportional representation electoral system, party system fragmentation in Portugal has remained relatively low and was, until recently, declining, a singular trend among established democracies (Lijphart, 2012: 75). Since 1985, cabinets have either been single-party governments (often with an absolute majority) or minimum-winning ideologically cohesive coalitions. The combination of persistently high salience of economic issues and high clarity of responsibility for outcomes makes Portugal a clearcut candidate for a solid and stable linkage between economic performance and the executive's public approval. However, several blind spots remain. We know, for example, that the relationship between objective economic performance and executive approval (Carlin et al., 2023) or support for incumbent parties (Freire and Santana-Pereira, 2012; Lobo and Pannico, 2020) has been stronger in some periods than in others in Portugal. Despite sharply contrasting economic conditions, consecutive elections have yielded similarly hefty punishments for government parties (Magalhães, 2017). Forecasting models of incumbent party vote have been required to incorporate political variables to account for the insufficient 10 A version of this Chapter was published as: Aguiar-Conraria, L., Fernandes, B., & Magalhães, P. C. (2023a). The economy and executive approval in a semipresidential regime: the case of Portugal. In T. Hellwig, & M. M. Singer (Eds.), Economics and politics revisited: Executive approval and the new calculus of support (pp. 108–130). Oxford University Press. 21 predictive power of economic variables (Magalhães and Aguiar-Conraria, 2009). In other words, the questions of whether vote or popularity functions are stable and of the variance in public support for the incumbent explained by the economy (the "e-fraction" of those functions), famously raised by Paldam (1991), remain relevant and intriguing for the case of Portugal. In this chapter, we advance two arguments about how to reconcile the evidence about the simultaneous importance and instability of economic performance as a driver of executive support in Portugal. The first points to the economic context itself and what it should imply for the clarity of performance signals available to voters. More specifically, we investigate whether executives are more punished by adverse economic developments than they are rewarded by positive ones. Although the possibility of such asymmetry in the relationship between economic performance and approval has been part of the literature since its modern inception (Mueller, 1970), the relatively short democratic electoral history of Portugal has contributed to leaving that hypothesis unexamined. To address that gap, we investigate whether the relationship between economic performance and executive approval in Portugal differs depending on whether the economy is expanding or contracting. The second relevant context, we argue, is political and institutional . In Portugal’s semi-presidential system, prime ministers who head the executive coexist with popularly elected and term-limited presidents who, while not sharing governing responsibilities, hold substantive powers and enjoy considerable public visibility. We suggest that the unstable nature of the political relationship between prime ministers and presidents in Portugal shapes the economics-approval link. To be sure, the literature on executive approval in semi-presidential systems has already suggested this is the case, treating periods of cohabitation — when the president and prime minister come from different parties or party coalitions — as causing a switch of responsibility from one incumbent to another, and therefore a switch in which actor voters hold accountable for the economy (Hellwig and Samuels, 2008; Elgie, 2018). We argue, however, that Portugal displays different features with different consequences. What matters in Portugal’s brand of semipresidentialism is not cohabitation per se but rather whether these periods are characterized by a conflictual relationship between prime ministers and presidents. In such conflictual contexts, we argue, clarity of responsibility for policy outcomes should be obscured, and the relationship between economic performance and the approval of both entities weakened. This calls attention to the importance of institutional factors behind the "instability dilemma" addressed in Carlin et al., 2023. Taken together, the two arguments help us understand some of the puzzling instability in the relationship between the economy and approval in Portugal, hopefully also contributing to the study of public support for prime ministers (and presidents) in semi-presidential systems similar to the Portuguese. 22 The chapter is structured as follows. Chapter 3.2 discusses asymmetry and semi-presidentialism in the literature on popularity functions. Chapters 3.3 and 3.4 describe the data and methodology used and the econometric results. Chapter 3.5 concludes. 3.2. Public approval and the economy in Portugal 3.2.1. Economic Contexts and Asymmetry The study of the relationship between economic performance and executive approval in Portugal has resulted in relatively few studies but with broadly consistent results. The pioneering work of Veiga (1998), using monthly approval data from 1986 to 1996, showed that prime ministerial popularity, besides receiving a boost immediately after the election (a honeymoon effect) and declining with time (cost or ruling), was negatively affected by increases in the unemployment rate. Later, revisiting a more extended time series of similar data, Veiga and Veiga (2004a and 2007) confirmed the previous findings. Other subsequent studies using aggregate data, this time focusing on the vote function, have confirmed the importance of economic performance — particularly of unemployment and GDP growth — as drivers of incumbent support (Veiga and Veiga, 2004b and 2010; Freire and Lobo, 2005; Magalhães and AguiarConraria, 2009; Goulart and Veiga, 2016). However, has economic performance been equally consequential at all times? Recently, Lobo and Pannico (2020) showed that "economic voting" in Portugal seems to have become stronger with the economic crisis of the early 2010s. Similarly, Carlin et al. (2023) show that the relationship between objective economic indicators and executive approval in Portugal became stronger in the period that encompasses the worst economic crisis experienced by the country in decades. This suggests that crises may render economic performance more politically consequential. In fact, from a cross-national perspective, Lewis-Beck and Nadeau (2012: 475) had already detected that economic variables had a greater impact on incumbent support in Southern than in Northern Europe. They concluded that "[t]he greater weight of the economic vote in the PIGS [Portugal, Italy, Greece, and Spain] countries appears almost entirely due to their poorer economic position, coupled with their less complex links between governance and policy execution." In sum, the economy seems to "matter more" in contexts where economic conditions are more adverse. Why should that occur? A well-known hypothesis points to an asymmetry in economic voting (Mueller, 1970; Nannestad and Paldam, 1997; Bélanger and Meguid, 2008; Dassonneville and Lewis- 23 Beck, 2014). As Evans and Andersen put it, economic downturns "elicit shared and reasonably perceptive responses that are not powerfully affected by political conditioning" (2006: 195). Under an economic crisis, the information conveyed by the mass media is more unequivocally negative than it is positive under economic good times (Soroka, 2006), and the salience of economic issues increases, boosting the relevance of economic performance for government approval (Singer, 2011). In contrast, during periods of growth, the economy loses importance, and the signals conveyed by the "mediated economy" become less unequivocal (Evans and Andersen, 2006:195; see also Chzhen et al. , 2014 and Dickerson, 2016). The 2011 and 2015 elections in Portugal illustrate this phenomenon. In the former, a major economic crisis led to large losses for the incumbent center-left Socialist Party (PS). However, in the latter, a clear economic recovery failed to prevent an almost equally large punishment for the ruling center-right coalition. Magalhães (2017) shows that while citizens' evaluations of the economy were fundamentally exogenous in relation to political predispositions during the period of economic crisis, in 2015, they were powerfully contaminated by partisanship. This suggests that the instability in the vote function in Portugal may partially result from how buoyant or stagnant economic conditions are differentially transmitted to incumbent support. To the extent that these general ideas about the vote function also apply to the popularity function, we should see the approval of the Portuguese prime minister suffering more under economic downturns than it is boosted by economic recoveries. In other words, we should expect the relationship between economic performance and prime ministerial approval to be asymmetric . 3.2.2. Portuguese Semi-Presidentialism and its Implications A second potential source of instability in the economics-approval link concerns the institutional and political context under which voters evaluate incumbents. The idea is far from new. The "clarity of responsibility" argument is predicated on the notion that the relationship between economic performance and public support depends on the extent to which there is a "perceived unified control of policymaking by the incumbent government" (Powell and Whitten, 1993: 398). In contexts where such perceived control is lacking — characterized by greater horizontal or vertical power-sharing — the economics-approval link should weaken. In such circumstances, it becomes more difficult to determine who should be retrospectively blamed or rewarded, and voters should discount economic performance as a signal of government competence (Duch and Stevenson, 2008). 24 In the study of executive approval in semi-presidential systems —— where a popularly elected head of state coexists with a prime minister accountable before parliament (Elgie, 1999) — the notion that context affects responsibility for the economy has also been present. However, informed by the experience of Fifth Republic France, the main focus of that inquiry has been how semi-presidentialism allows for two alternating patterns of economic accountability. Under a unified government, when French presidents are supported by a parliamentary majority of their own party, they become the leaders of that majority and thus of the executive, giving the system an unmistakable presidential dynamic (Shugart, 2005). In contrast, under "cohabitation," when presidents and prime ministers belonged to different parties, power shifts to the latter, creating a "parliamentary" dynamic. The presence or absence of cohabitation was thought to determine an oscillation between phases when the system became de facto , respectively, parliamentary and presidential (Duverger, 1980; Lijphart, 2012). This, in turn, generated another oscillation: on whom voters held accountable for economic performance. In the presidential phases, under a unified government, the French president's approval was strongly affected by economic performance. However, in the parliamentary phases, under cohabitation, the responsibility for economic governance shifted to the prime minister. During these periods, the economy’s influence on presidential support dropped significantly, while the opposite occurred with prime ministerial approval (Lewis-Beck, 1997; Lewis-Beck and Nadeau, 2004). 11 In 2002, the synchronization of presidential and legislative elections made cohabitation in France much more unlikely, ending this oscillation. However, the French experience still molds the thinking about economic accountability in semi-presidential systems, with studies finding the same kind of oscillating pattern in countries other than France. Looking at election results in many semi-presidential systems, Hellwig and Samuels (2008) showed that presidents are rewarded and punished for the economy only under conditions of unified government, while the PM's party is only rewarded and punished under conditions of cohabitation. Elgie (2018), finding similar results, suggested the existence, in the study of semi-presidential systems, of "a basic scientific consensus about the effect of cohabitation on economic voting" (Elgie, 2018:103). However, we propose that this pattern is unlikely to extend to a semi-presidential system like Portugal’s. Since 1986, all presidents in Portugal have been prominent political figures, all former leaders of their parties. Presidents enjoy important powers, including the possibility of vetoing legislation or referring bills or statutes for constitutional review. Presidents also employ "going public” tactics, using the authority and prestige of their office to direct public attention to the issues they prioritize (Neto and 11 But see also Turgeon et al. (2015) showing that only prime ministerial approval is affected by economic variables, and only during cohabitation. 25 Lobo, 2009; Jalali, 2011). And they even have the (mostly) unconstrained possibility of dissolving parliament in the absence of a prime ministerial resignation or a successful motion of censure. This is a power they have used with parsimony in the last four decades, but which nevertheless has been used and stands as a relevant shadow over governments. Yet, while never becoming purely parliamentary, the Portuguese system's dynamic has never become presidential either, de jure or de facto . Portuguese presidents are not the leaders of their parties. The head of the executive is always the prime minister, i.e., the party and parliamentary majority leader with whom presidents coexist. Accordingly, presidential elections in Portugal bear some traits of a "second-order" election (Reif and Schmitt, 1980), most notably their lower level of turnout in comparison with legislative elections (Fortes and Magalhães, 2005; Franco, 2020). The implication is that, at least in Portugal, there has been no oscillation between parliamentary and presidential phases depending on cohabitation or unified government. This, in turn, suggests that an oscillation in how voters hold prime ministers and presidents accountable for the economy should be absent as well. However, this does not mean that the relationship between presidents and prime ministers is irrelevant to the role of the economy in the V-P function. Although it does not change who runs the everyday operation of government, cohabitation has nevertheless been a source of political conflict between presidents and prime ministers under semi-presidentialism (Protsyk, 2005a and 2005b; Elgie, 2010). The existing literature provides clear evidence about the conditions that have favored such conflictual relationship in Portugal. Portuguese presidents can only serve two consecutive terms of office. First-term incumbents who seek reelection for a second term in office — so far, all of them — have incentives to preserve a non-conflictual relationship with prime ministers and their parliamentary majorities in order to secure the required absolute majority of the vote in their reelection bid. This has occurred even in contexts of cohabitation. Since 1986, presidents Mário Soares, Jorge Sampaio, Cavaco Silva, and Marcelo Rebelo de Sousa successfully avoided major confrontations with prime ministers during their first term, even when they belonged to different parties. By defusing the potential for conflict, some incumbent presidents have even been able to secure opposing governing parties' explicit or implicit support in their reelection bids, as in the case of the Socialist Mário Soares in 1991 or PSD’s Marcelo Rebelo de Sousa in 2021. However, in their second term, the combination of cohabitation and the lack of a reelection incentive has typically led presidents to engage in a more conflictual relationship with prime ministers. As Jalali (2011: 170) puts it, while presidents' first terms are characterized by a "considerably weaker activism (…), not least motivated by the need to maintain a majority for reelection," second terms typically 26 show a rise in “legislative interventionism” and “guerrilla warfare” on the part of presidents. In fact, second terms have only been historically peaceful when presidents face a government majority congruent with them in partisan terms (Freire and Santana-Pereira, 2019). However, under cohabitation, presidents in their second and last term of office have used their veto and constitutional referral powers more frequently and have engaged in more open political confrontations with the government (Magalhães, 2001; Neto and Lobo, 2009; Jalali, 2011; Franco 2020). This suggests that, by favoring inter-institutional conflict between prime ministers and presidents, second-term cohabitations should contribute to blurring accountability for the economy. By fostering “legislative interventionism” on the part of presidents and “constraining the governments’ capacity for political leadership” (Jalali, 2011: 169), second-term cohabitations are periods when what Powell and Whitten defined as "clarity of responsibility” — the "perceived unified control of policymaking by the incumbent government" (1993: 398) — is undermined. This leads us to expect that the relationship between economic performance and prime ministerial approval, as we hypothesized earlier, should be weaker when the president is of a different party and is serving in their second and last term of office. In summary, we propose two main hypotheses. H1 is that the relationship between economic performance and prime ministerial approval should be contingent upon the economic context itself: stronger under crises and weaker under expansions (i.e., asymmetric). H2 is that it should also be contingent upon the relationship between prime ministers and presidents: stronger under peaceful coexistence, weaker under contexts that have favored political conflicts, i.e., during (presidential) secondterm cohabitations. 3.3. Data, Variables, and Estimation Strategy 3.3.1. Data and Variables Our dataset encompasses two measures of approval and a set of economic and political variables. We extracted two approval ratings from the Executive Approval Project dataset (Carlin et al. , 2019). Our dependent variables are smoothed versions of the quarterly approval ratings of the Portuguese president and prime minister ( Presidential Approval Rating and PM Approval Rating ). The data available in the Executive Approval Database 2.0 goes from the second quarter of 1986 to the third quarter of 2018. Table 5 lists the prime ministers and presidents throughout this period and indicates whether presidents were serving their first or second term and whether they and the prime ministers belonged to 27 the same or different parties. The latter two aspects are crucial to test our second hypothesis and show considerable diversity through time. Portugal has had five presidents since the promulgation of the current 1976 democratic constitution. Since the approval series began in 1986, the first president, Ramalho Eanes, is excluded from our analysis. The last we consider is Marcelo Rebelo de Sousa, who was elected in 2016. At the time of this writing, all presidents spent ten years in office, 12 serving two consecutive terms of five years. Regarding the executive branch, headed by the prime minister, several personalities have been in power since 1986. Only two parties occupied the prime ministerial post during this period: the PS and PSD. Another party, CDS-PP, has integrated some of PSD's governments to achieve a parliamentary majority. Table 5: Portuguese Presidents and Prime Ministers between 1986 and 2018. Figure 5 shows the quarterly approval rates of Portuguese prime ministers (black line) and presidents (grey line). Shaded areas represent periods of cohabitation, either during the president’s first 12 The exception is the current president, Marcelo Rebelo de Sousa, who was elected in 2016 and reelected in 2021. Period Prime Minister (Party) President (Party) Presidential Term Cohabitation 1986-1991 Cavaco Silva (PSD) Mário Soares (PS) First Yes 1991-1995 Cavaco Silva (PSD) Mário Soares (PS) Second Yes 1995-1996 António Guterres (PS) Mário Soares (PS) Second No 1996-2001 António Guterres (PS) Jorge Sampaio (PS) First No 2001-2002 António Guterres (PS) Jorge Sampaio (PS) Second No 2002-2004 Durão Barroso (PSD) Jorge Sampaio (PS) Second Yes 2004-2005 Pedro Santana Lopes (PSD) Jorge Sampaio (PS) Second Yes 2005-2006 José Sócrates (PS) Jorge Sampaio (PS) Second No 2006-2011 José Sócrates (PS) Cavaco Silva (PSD) First Yes 2011 José Sócrates (PS) Cavaco Silva (PSD) Second Yes 2011-2015 Pedro Passos Coelho (PSD) Cavaco Silva (PSD) Second No 2015-2016 António Costa (PS) Cavaco Silva (PSD) Second Yes 2016-2018 António Costa (PS) Marcelo Rebelo de Sousa (PSD) First Yes 28 (light-grey) or second term (dark-grey). The data covers four presidencies and ten prime ministerial cabinets. With brief exceptions in the early 1990s and late 2015, Portuguese presidents have been more popular than prime ministers throughout the entire period. While visual inspection suggests the two series are related, the relationship is clearly imperfect, and there are even periods where trends seem to diverge, suggesting that they may be driven by somewhat different processes. Table 6 provides descriptive statistics for both the Prime Ministerial and Presidential approval ratings. Table 6: Descriptive statistics of the variables employed. Observations Mean Std Dev Min Max PM Approval Rating 130 39.80 10.41 18.54 66.13 Presidential Approval Rating 130 58.59 13.06 32.18 82.35 GDP Growth 130 2.20 2.83 -4.53 8.71 GDP Growth Negative 130 0.40 1.03 0.00 4.53 GDP Growth Positive 130 2.60 2.21 0.00 8.71 Honeymoon PM 130 0.21 0.55 0.00 2.00 Honeymoon President 130 0.16 0.49 0.00 2.00 Cohabitation 1st term 130 0.38 0.49 0.00 1.00 Cohabitation 2nd term 130 0.25 0.43 0.00 1.00 Unified government 130 0.37 0.48 0.00 1.00 Figure 5: Prime Ministerial and Presidential Approval Rating, quarterly data. Note: Data from Carlin et al. (2019). 35 Table 8: Presidential and prime ministerial approval, the economy, and cohabitation. While Models 2 and 3 showed that GDP Growth Negative t-1 was harmful to the Prime Minister, Model 4 shows that such negative value results from the aggregation of three coefficients for three different contexts. Cohabitation 1st term * GDP Growth Negative t-1 and Unified government* GDP Growth Model 3 Model 4 PM Approval Presidential Approval PM Approval Presidential Approval PM Approval Rating t-1 0.769*** (0.038) - 0.771*** (0.038) - Presidential Approval Rating t-1 - 0.769*** (0.054) - 0.773*** (0.053) Honeymoon PM 3.038** (0.920) 3.073** (0.909) - Honeymoon President _ 1.638 (1.869) - 1.634 (1.868) Cohabitation 1st term 1.940† (1.037) 3.337* (1.483) 1.998† (1.109) 3.476† (1.796) Cohabitation 2nd term 0.069 (1.057) 1.675 (1.372) -0.356 (1.095) 1.889 (1.558) GDP Growth Negative t-1 -1.066*** (0.237) - - - Cohabitation 1st term* GDP Growth Negative t-1 - - -1.331** (0.449) - Cohabitation 2nd term * GDP Growth Negative t1 - - 0.511 (0.819) - Unified government* GDP Growth Negative t-1 - - -1.138*** (0.272) - GDP Growtht-1 - 0.596** (0.184) Cohabitation 1st term* GDP Growtht-1 - - - 0.566* (0.264) Cohabitation 2nd term* GDP Growtht-1 - - - 0.442 (0.664) Unified government * GDP Growtht-1 - - - 0.651** (0.250) Constant 8.112*** (1.558) 10.529*** (2.815) 8.044*** (1.591) 10.250*** (2.676) Var (e.PM Approval) 19.589 (3.108) 19.046 (3.114) Var (e.Pres Approval) 34.984 (6.240) 35.091 (6.232) Cov(e.PM Approval, e.Pres Approval) 13.899 (3.074) 13.567 (3.055) Observations 129 129 SRMR 0.013 0.013 Note: Standard-errors between brackets. Statistical significance: †p<.10; *p<.05; **p<.01; ***p<.001 (two-tailed tests). 36 Negative t-1 are both negative, significant, and similarly sized. In other words, economic performance affects the popularity of prime ministers under conditions of first-term cohabitation or unified government, and it does so asymmetrically, as posed in H1. In contrast, the coefficient for the interaction Cohabitation 2nd term * GDP Growth Negative t-1 for prime ministers is far from statistical significance at conventional levels. In other words, under conditions that have been characterized by a more conflictual relationship between prime ministers and presidents, prime ministerial approval becomes insensitive to economic performance. Similarly, Models 2 and 3 showed the coefficient GDP Growth t-1 is positive for the president. Model 4 shows these results from Cohabitation 1st term * GDP Growth t-1 and Unified government* GDP Growth t-1 , whose coefficients are positive, significant, and similarly sized: under conditions that favor peaceful coexistence with prime ministers, presidential approval is (symmetrically) affected by economic performance. However, like what occurs with prime ministerial approval, the interaction Cohabitation 2nd term * GDP Growth t-1 for presidents is not statistically significant. 21 In sum, contextual conditions that combine a lack of reelection incentives for presidents and partisan divergence between them and prime ministers have been inimical to economic accountability: interinstitutional conflict between presidents and prime ministers makes the popularity of both entities less sensitive to economic performance, particularly in the case of the latter. 3.5. Conclusion It has long been suspected that support for incumbent governments and parties in semi-presidential systems might be characterized by somewhat different dynamics from those observed in pure parliamentary or presidential systems. In the well-studied French case, cohabitation seemed to switch responsibility for the economy from the president to the prime minister, as the leadership of the executive moved from the former to the latter (Lewis-Beck, 1997). Other studies have even suggested this also occurs with semi-presidential systems more broadly, with cohabitation serving as the switch that makes presidents (under the former) or prime ministers and their parties (under the latter) publicly accountable for the economy. 21 However, it has the same sign as the other interactions, suggesting the contrast between 2nd term cohabitations and the remaining situations is more robust for the case of the prime minister than for the case of the president. 37 We suggest, however, that this dynamic has not been reproduced in the case of semi-presidential Portugal, or at least not since the mid-1980s. Because cohabitation in Portugal does not switch the locus of responsibility for economic policy, it also does not switch the locus of responsibility for economic performance from the public’s point of view either. Voters strongly punish prime ministers for bad economic performance under either cohabitation or unified government. There is, however, one crucial exception: conflictual cohabitations , such as those that have prompted presidents to make more extensive uses of their legislative and oversight powers to place important obstacles to governments' policy agenda during presidential second (and last) terms. Under such conditions, public approval for both entities, prime ministers and presidents, has been mostly insensitive to economic performance. These findings are compatible with the notion that conflicts between politically relevant presidents and the prime ministers they oppose contribute to blurring responsibility for the economy. By doing so, the study findings point to a source of instability in the Portuguese V-P function that, while rooted in institutional factors — semi-presidentialism, presidential powers, term-limits — is far from static: it depends on the political context shaping the relationship between prime ministers, presidents, and their parties, a context that has varied significantly throughout the period under examination (see Table 5). Furthermore, we also show that although the approval of both prime ministers and presidents is affected by the economy outside of second-term cohabitations, that relationship is asymmetric for prime ministers and symmetric for presidents. Future studies may be able to establish with greater certainty the reasons for that difference. However, we speculate that it is consistent with the different roles played by both entities in the political system. As partisan entities in charge of the executive, prime ministers are affected by the economy in a way that is compatible with the lower partisan contamination of economic evaluations in bad times (Evans and Andersen, 2006; see Magalhães, 2017, for Portugal), i.e., asymmetrically. Conversely, the relationship between the economy and presidential approval is less contaminated by such process, given the lower relevance of partisanship as a cue with which to evaluate presidents. The asymmetry found for prime ministerial approval addresses part of the puzzle concerning the apparent instability of the V-P function in the country (Freire and Santana-Pereira, 2012; Carlin et al., 2023). In particular, it helps accounting for the very strong connection between economic performance and prime ministerial approval during Portugal’s worst economic times since the 1970s. More broadly, this chapter makes two main contributions. First, a clearer understanding of some of the reasons why, in the case of Portugal, the link between economic performance and executive support appears unstable if approached in a way that ignores the variability introduced by economic and institutional contexts. A second contribution consists of potentially important clues to be explored in the 38 study of semi-presidential systems that bear similarities to the Portuguese. In countries such as Bulgaria, Poland, Lithuania, Romania, Slovakia, or Slovenia, for example, popularly elected presidents are neither like the French heads of the executive nor like the mostly ceremonial figureheads that can be found in the semi-presidentialism of Austria, Ireland, or Iceland. Their powers are often relevant, and their cohabitation with prime ministers who belong to different parties has been characterized by peaceful coexistence in some cases but conflict and even policy stalemates in others (Protsyk, 2005a and 2005b; Elgie, 2010). The Portuguese case provides a few testable hypotheses about how economic performance should affect the approval of both prime ministers and presidents in such contexts. 39 4. US Presidential Approval Rating: Time-Frequency Approach 22 4.1. Introduction One of the challenges posed by the study of popularity functions has been the difficulty in distinguishing between structural and transient effects on government approval of different economic variables while at the same time addressing issues of non-stationarity and non-linearity. One of the consequences of this difficulty has been a plethora of inconsistent findings, even in the case where the relationship between economic factors and government approval has been most exhaustively studied: the United States (Berlemann and Enkelmann, 2014). In this work, we show how wavelet analysis can be employed in conjunction with other econometric methods to address these problems. We describe the advantages of wavelet tools, particularly how they allow us to decompose time series data into a sum of waves of different frequencies, provide local analyses unaffected by non-stationarity and non-linearity problems, and, with advancements (Aguiar-Conraria and Soares, 2014; Aguiar-Conraria et al. , 2023), perform multivariate analysis. To show how the use of multivariate wavelet analysis allows for a more nuanced understanding of the relationship between the economy and presidential approval, we illustrate this approach, focusing on the approval ratings of President Barack Obama and comparing the results with OLS estimations. To analyze the relationship between the economy and the approval, we utilized the conventional economic variables: unemployment and inflation rates. We also used consumer sentiment as a measure of economic perception. As our focus is President Obama’s presidency, our data goes from January 2009 to December 2016. The results reveal how Obama’s approval rating was structurally influenced by economic variables such as unemployment and inflation, with different synchronizations, but also how they were transiently related to consumer sentiment, in a result that would not emerge from traditional econometric models. 22 A version of this chapter has already been submitted to a academic journal. 40 The study is organized as follows. In Chapters 4.2 and 4.3, we provide a self-contained summary of wavelet analysis, describing the advantages of the tools we use and why they are relevant to the study of popularity functions and complementing with an example. In Chapter 4.4, we describe the data that will be employed to illustrate the potential of wavelet analysis to this topic. Chapter 4.5 presents the results of both a traditional OLS approach and wavelet analysis, highlighting the added value of the latter. Chapter 4.6 concludes. 4.2. Wavelet analysis and the study of popularity functions 4.2.1. The challenge of instability Ever since Mueller (1970), a profusion of studies has empirically investigated how the economy affects presidential approval in the United States. While several indicators have been used as proxies for assessing economic conditions, unemployment and inflation rates have been among the most employed in the V-P function literature since the 1970s. Voters are assumed to recognize that the state of the economy is unfavorable when unemployment is high, and that high inflation acts as a burden on their income, and they should use that information to assess the performance of incumbents. However, contrary to initial expectations, the correlation between economic performance and executive popularity has been less than consistent. This instability can be attributed to many factors. One possibility is the disparate reactions elicited by leftand right-wing incumbents. For example, left-leaning executives may experience intensified criticism during periods of heightened unemployment, whereas their right-leaning counterparts face amplified scrutiny amidst inflation (Hibbs, 1977; Powell and Whitten, 1993; Swank, 1998). Moreover, the responses to economic shocks are not symmetric. As Soroka (2006) noted, adverse economic developments typically result in a sharper decline in popularity compared to the mild uptick associated with positive economic surprises. However, yet another possibility is that these effects may be more or less transitory. For example, events may lead voters to set aside economic and political preferences temporarily. In the V-P function literature, events such as wars or terrorist attacks elicit a well-known phenomenon referred to as the “rally ‘round the flag” effect (Mueller, 1970; Newman and Forcehimes, 2010), wherein individuals unite in support of their political leadership, blurring the relation between economic fundamentals and approval ratings. For instance, despite the American economy facing an economic recession in 2001, President George W. Bush’s popularity experienced a significant surge following the 9/11 terrorist attack. 41 More generally, as Gronke and Newman (2003) highlight, presidential performance evaluations are likely to be inherently unstable. Based on the media priming literature, they argue that media emphasis on specific issues shapes the criteria citizens employ to evaluate presidential performance. If public expectations of presidents are dynamic and evolve over time, a single statistical model cannot universally capture the drivers of approval ratings. Accordingly, they advocate for the development of new estimation methods that accommodate time-varying coefficients. Similarly, Berlemann and Enkelmann (2014), while suggesting that inflation and unemployment seem to exert a long-term influence on presidential approval in the U.S., show inconsistencies in short-term effects, attributing them as well to a potentially evolving relationship between economic factors and popularity. Furthermore, they also question both the prevalent use of linear estimation in previous studies, proposing that nonlinear relationship might exist, as well as the fact that many studies have not adequately considered stationarity issues. In fact, Berlemann et al. (2015) and Choi et al. (2016) find an interdependent and non-linear impact of the unemployment and inflation rates on US presidential popularity. These different insights serve as an invitation to take advantage of the possibilities awarded by wavelet analysis. Ever since Fourier's time, we have known that many time series can be expressed as a sum of periodic cycles. This has laid the foundation for methods like the Fourier transform and wavelet transform, which enable us to decompose time series data into a sum of sinusoidal waves of varying frequencies, offering insights into the data's periodic patterns. The wavelet transform has, however, a crucial advantage over the Fourier transform: it allows for patterns in the data and in the relationships between variables to be time-varying . Accordingly, as a local analysis, it assumes neither a time-invariant generating process (stationarity) nor that relationships between variables are linear. In the following Subchapters, we underscore the benefits of employing continuous wavelet analysis in conjunction with traditional econometric methods to address the challenge posed by instability. 4.2.2. The (partial) wavelet coherency, gain, and phase-difference Earlier studies have already delineated the merits of wavelet analysis for political science (Aguiar-Conraria et al. , 2012 and 2013). More recently, we find some applications of the bivariate analysis in political science (Wang, 2023a and 2023b). Aguiar-Conraria et al. (2014) went a step further by allowing an analysis involving more than two series – multivariate wavelet analysis. This methodology has been applied in different topics, such as macroeconomics and, to a less extent, in political science (Wang, 42 2023a). Recently, some studies have estimated equations in the time-frequency domain. Aguiar-Conraria et al. (2023) provide a recent application, though its utilization in political science remains unexplored. The studies primarily utilize three wavelet tools: the wavelet power spectrum, (partial) wavelet coherency, and (partial) wavelet phase-difference. The wavelet power spectrum estimates a variable's variance in the time-frequency domain, pinpointing the contribution of specific frequencies to total variance and identifying significant cycles. Wavelet coherency, akin to the absolute value of the correlation coefficient, estimates the strength of the association between two variables across different times and frequencies, revealing varying correlations. Phase-difference informs the sequence/synchronization and nature (positive/negative) of variable correlations. The extensions, partial wavelet coherency, akin to the partial correlation coefficient, and partial phase-difference, provide the same information while accounting for third variables. Finally, the multiple wavelet coherency: its interpretation is similar to the R-squared but, of course, done in the time-frequency domain. In particular, the partial phase-difference allows us to distinguish between positive and negative correlation, after taking into account the effects of a third variable, and it also informs us on which variable is leading which. Let 𝜙𝑥,𝑦|𝑧 be the phase-difference between 𝑌𝑡 and 𝑋𝑡 after controlling for 𝑍𝑡. If 𝜙𝑦,𝑥|𝑧 =0, then the two variables’ series move together at a given time and frequency. If 𝜙𝑦,𝑥|𝑧 є (0,𝜋 2) then they are in phase (positive correlation) with 𝑌𝑡 leading 𝑋𝑡, and if 𝜙𝑦,𝑥|𝑧 є (−𝜋 2,0), the series are in phase, with 𝑋𝑡 leading. On the other hand, if 𝜙𝑦,𝑥|𝑧 є (−𝜋,−𝜋 2), series 𝑌𝑡 leads 𝑋𝑡 in an out of phase relation (negative correlation), and if 𝜙𝑦,𝑥|𝑧 є (𝜋 2,𝜋) the series 𝑌𝑡 and 𝑋𝑡 are out of phase, with 𝑋𝑡 leading. Figure 7 summarizes this information in a circle scheme. Figure 7: Phase-Difference circle. Note: Aguiar-Conraria and Soares (2014). 43 Although highly beneficial, the wavelet tools described previously could not perform regressions in the time-frequency domain. This limitation was addressed by Aguiar-Conraria et al. (2023) who extended the concept of wavelet gain, proposed by Mandler and Scharnagl (2014), to a multivariate context. They demonstrated that the partial wavelet gain of 𝑌𝑡 over 𝑋𝑡, while controlling for 𝑍𝑡, and 𝑌𝑡 over 𝑍𝑡, while controlling for 𝑋𝑡, can be interpreted as the coefficient in the linear regression of 𝑌𝑡 on the explanatory variables 𝑋𝑡 and 𝑍𝑡 at each specific time and frequency. In the following Chapter, we provide an intuitive explanation, illustrated by a specific example, of interpreting graphical representations and extracting pertinent information related to partial coherency, partial phase-difference, and partial wavelet gain. For the formulae and rigorous mathematical treatment, the reader is directed to Aguiar-Conraria et al. (2012) and Aguiar-Conraria and Soares (2014). 4.3. An example 4.3.1. The Data Generating Process Suppose a variable 𝑋𝑡 is primarily composed of two periodic cycles: a one-year cycle and a three-year cycle, along with some white noise. Assume we have 10 years of monthly data. In this case, the expression for 𝑋𝑡 can be written as: 𝑋𝑡=cos(2𝜋𝑡 1)+cos(2𝜋𝑡 3)+𝜀𝑥,𝑡, 𝑓𝑜𝑟 𝑡=0, 1 12,2 12,…,10. 23 Assuming still that we have monthly data spanning 10 years, consider another variable, 𝑍𝑡, which comprises both white noise and a six-month cycle, expressed as: 𝑍𝑡=cos(2𝜋𝑡 1 2 ⁄)+𝜀𝑧,𝑡, 𝑓𝑜𝑟 𝑡=0, 1 12,2 12,…,10. Finally, consider a variable 𝑌𝑡 that is a function of the previous two. However, to fully appreciate the power of wavelet estimation, let's introduce some complexities. First, the relationship between 𝑋𝑡 and 𝑌𝑡 is not constant either across time or frequencies. For example, since it is entirely possible for one 23 The use of cosines might initially appear unusual; however, Aguiar-Conraria et al. (2012) demonstrate that several autoregressive processes, as estimated in the political science literature, can indeed be characterized in this manner. 44 variable to have a positive effect on another in the short term and deleterious effects in the long term, we assume that the relationship between 𝑋𝑡 and 𝑌𝑡 is positive at higher frequencies (corresponding to the one-year cycle in our example), while negative at lower frequencies (the three-year cycle). Second, to encapsulate time-varying relationships, we modify the regression coefficient that connects the one-year cycle between 𝑋𝑡 and 𝑌𝑡. It is set at 1 for the first five years of data and then adjusts to 3 for the remaining period. Lastly, to address transient relationships, we propose that the influence of 𝑍𝑡 on 𝑌𝑡 is temporary, lasting only for one year, specifically when 2≤𝑡<3. Putting all this information together, we have that 𝑌𝑡=𝛼1,𝑡cos(2𝜋𝑡 1 2 ⁄)+𝛼2,𝑡cos(2𝜋(𝑡−1 12 ⁄ ) 1)+𝛼3,𝑡cos(2𝜋(𝑡−3 12 ⁄ ) 3) +𝜀𝑦,𝑡 𝑓𝑜𝑟 𝑡=0, 1 12,2 12,…,10. With 𝛼1,𝑡 =1 if 2≤𝑡<3 and zero, otherwise; 𝛼2,𝑡 =1 if 𝑡≤5 and 𝛼2,𝑡 =3 if 5<𝑡≤10; 𝛼3,𝑡 =−2. We've added another twist to consider: at higher frequencies, there is a one-month lag for the impact of 𝑋𝑡 on 𝑌𝑡, and at the three-year frequency, the lag extends to three months. However, the influence of 𝑍𝑡 on 𝑌𝑡 is instantaneous. 4.3.2. Reading the results The results of the wavelet analysis of this example are presented in Figure 8. On the top, we have the wavelet multiple coherency (chart a). Still on the left, we have the partial wavelet coherency between 𝑌𝑡 and 𝑋𝑡, after controlling for 𝑍𝑡 (chart c.i), and between 𝑌𝑡 and 𝑍𝑡, after controlling for 𝑋𝑡 (chart d.i). The color code for coherency (partial or multiple) ranges from blue (low coherency – close to zero) to red (high coherency – close to one). The black contours indicate the 5% significance level for the coherency computed with 5000 Monte-Carlo simulations. A parabola-like black line indicates the region affected by edge effects. Results should be interpreted with care in this area. 24 24 As with other types of transforms, the Continuous Wavelet Transform applied to a finite length time series suffers from border distortions due to the fact that the values of the transform at the beginning and the end of the time series involve missing values of the series, which are artificially prescribed. One usually pads the series with zeros to avoid wrapping. These edge-effects increase for lower frequencies. In this area of the time-frequency plane, the results are subject to border distortions and have to be interpreted carefully. 51 unemployment and inflation. Then, we expand the model to include consumer sentiment. The models were estimated using the Ordinary Least Squares. We first estimate a static model. Given that our variables do not exhibit a unit root, 26 we cannot say that it corresponds to a cointegration relation, but it still captures the idea of a long-run relation: 𝑂𝑏𝑎𝑚𝑎 𝐴𝑝𝑝𝑟𝑜𝑣𝑎𝑙𝑡=𝛼+ 𝛾 𝑿′𝑡−1+𝛿 𝒁′𝑡+𝜀𝑡. Second, we add dynamics to the model by including the lagged dependent variable as a regressor. 𝑂𝑏𝑎𝑚𝑎 𝐴𝑝𝑝𝑟𝑜𝑣𝑎𝑙𝑡=𝛼+ 𝛽 𝑂𝑏𝑎𝑚𝑎 𝐴𝑝𝑝𝑟𝑜𝑣𝑎𝑙𝑡−1+ 𝛾 𝑿′𝑡−1+ 𝛿 𝒁′𝑡+𝜀𝑡. In the above equations, 𝑂𝑏𝑎𝑚𝑎 𝐴𝑝𝑝𝑟𝑜𝑣𝑎𝑙𝑡−1 is the lagged dependent variable, 𝑿′𝑡−1 is a vector of independent economic variables, and 𝜀𝑡 is the error term. To address political factors (𝒁′𝑡), we incorporated two extra variables in the econometric model: a "honeymoon effect" with values 2 for the initial month, 1 for the second, and 0 thereafter, to factor in the initial popularity surge post-presidential inauguration (first and second term); and a "first term" dummy variable, to account for Obama's initial presidential mandate and for erosion of popularity. We expected both to have positive coefficients. The results are presented in Table 9. Upon initial observation, it seems evident that the economy exerted an impact on Obama’s popularity. The long-run equations, columns (1) and (2), show that both the unemployment rate and the inflation rate have strong, significant, and negative effects on the approval rating. However, the Consumer Sentiment (column 2) is completely irrelevant (p-value of 0.94). 27 When one considers a dynamic model able to capture the short-run dynamics, columns (3) and (4), the results change somewhat. The inflation rate is no longer significant, and the consumer sentiment becomes highly significant. Subsequently, given these results, a researcher would probably estimate an Error Correction Model combining both equations, explicitly modeling the distinction between the longrun relation and the short-run dynamics. We do not perform that analysis here, as our purpose is a different one: to illustrate the advantages of complementing the traditional regression techniques with wavelet analysis. 26 Using the Augmented Dickey-Fuller test to assess the stationarity of the dependent variable, the test statistic for the approval rate is -3.295, which follows between the 1% critical values of -3.517 and the 5% critical value of -2.894. 27 Some authors can argue that, to estimate a long-run relationship, unemployment and inflation should not be lagged. Including the contemporary values has minimal impact on our estimations. We kept them lagged just for consistency with columns 3 and 4. 52 Table 9: Obama’s approval rate as a function of economic and political variables, estimated by OLS. (1) (2) (3) (4) L.Approval 0.878*** 0.890*** (0.041) (0.041) L.Unemployment -1.405*** -1.381** -0.697*** -0.382** (0.493) (0.599) (0.159) (0.172) L.Inflation -2.311*** -2.309*** -0.044 0.018 (0.377) (0.393) (0.189) (0.177) Sentiment 0.007 0.083*** (0.090) (0.029) Honeymoon 6.408*** 6.446*** 0.368 0.755* (1.160) (1.281) (0.520) (0.444) First_Term 8.014*** 8.049*** 2.119*** 2.472*** (1.664) (1.643) (0.634) (0.645) Constant 57.25*** 56.53*** 9.927*** 0.192 (3.058) (10.50) (2.406) (3.567) Observations 95 95 95 95 R 2 0.466 0.466 0.914 0.921 F statistic 17.74 15.43 167.920 269.229 Note: Standard-errors between brackets. Statistical significance: * p<.10; ** p<.05; *** p<.01 (two-tailed tests). In short, Table 9 tells us that the economy exerted an impact on Obama’s popularity. Clearly, the unemployment rate is relevant throughout the entire period under consideration. The inflation rate and the consumer sentiment are also relevant, but they are model dependent. It is also evident that the coefficients of our political variables are positive, as we expected. However, regarding statistical inference, the honeymoon effect is model dependent as well. The econometric models employed in our analysis did not provide additional information regarding the specific time periods and frequencies at which the relationship between approval rating and economic variables attained statistical significance. Subsequently, we will examine how these variables interact over time and frequency using wavelet analysis. 53 4.5.2. Wavelet Analysis We start with an examination of the Partial Wavelet Coherency, depicted in Figure 10. Clearly, unemployment (chart a.i) and inflation (chart b.i) rates exhibit a significant correlation with approval, particularly at low frequencies. During Obama’s tenure, a co-movement between his approval rating and the state of the economy is observed for frequencies between 2 years to 4 years. This finding aligns with our previous discussion, emphasizing the substantial influence of economic factors on Obama’s approval. The partial phase-differences in the middle of Figure 10 provide evidence for a negative correlation between Obama’s approval rating and both the unemployment rate and the inflation rate. Starting with the unemployment rate, we observe a substantial and statistically significant coherence during a large part of Obama’s presidency, particularly within the frequency band of 2-4 years. At those frequencies, the phase-difference (diagram a.ii) consistently falls between π/2 and π, indicating Figure 10: Wavelet Analysis: Obama’s Approval, Unemployment Rate, and Inflation Rate. Note: Top: Relation between the Approval rate and the Unemployment rate after controlling for Inflation. Bottom: Relation between the Approval rate and the Inflation rate after controlling for Unemployment. Charts a.i and b.i: Partial Wavelet Coherency between the Approval rate and one economic variable after controlling for the other. Diagrams a.ii and b.ii: Partial Phase-Difference between the Approval rate and one economic variable after controlling for the other in two frequency bands. Diagrams a.iii and b.iii: Partial Wavelet Gain between the Approval rate and one economic variable after controlling for the other in two frequency bands. 54 an inverse relationship between the two variables, with unemployment leading the way. 28 The phasedifference at higher frequencies conveys a similar message; however, given the low significance of the coherency, we refrain from attaching importance to this result. As we can see on the diagram a.iii, the estimated coefficient for the 2-4 years frequency band remains stable, hovering around 5. Regarding the inflation rate (depicted in chart b.i), we observe a similar frequency band of significant coherence (2-4 years), which persists throughout almost the entirety of Obama’s presidency. Within this frequency band, the partial phase-difference closely approximates to π (or –π), indicating a negative correlation between the inflation rate and the approval rate. Like the relationship observed between the unemployment rate and the approval rate, the inflation rate leads the approval rate. Instead, simultaneous increases (decreases) in the inflation rate coincide with corresponding decreases (increases) in the approval rate. The estimated coefficient for this frequency band fluctuates between 3.5 and 4. It is noteworthy that at higher frequencies, there is a small island of high coherency between 2015 and 2016. For that period, the phase-difference indicates a negative relationship with inflation leading again. In Figure 11, we add consumer sentiment to the analysis. Our results regarding inflation and unemployment do not change substantively when we use consumer sentiment as a control variable. The most important regions of high coherency occur at lower frequencies, and the phase-differences and estimated coefficients are similar. Visually, the most important change is about the area of the high coherency regions, which is now smaller. But that is to be expected, given the correlation between consumer sentiment and the economic variables. A notable finding emerges when one looks at the partial coherency between the approval rate and the consumer sentiment in Chart C of Figure 11, which helps to make sense of the OLS results in the previous Subchapter. Consumer sentiment is significant at high coherencies for very specific periods. That is particularly obvious between 2011 and 2012. After 2014, and at high frequencies, although not as high, it becomes significant again. This suggests that the state of the economy is of greater importance in the medium to long term, while the consumer sentiment captures more immediate effects. In contrast to the relationship established with unemployment and inflation, the connection between consumer sentiment and approval rating appears to be less stable. 28 See Figure 7. 55 Analyzing the diagram c.ii (primarily for higher frequencies, which exhibit statistical significance), we find that between 2011 and 2012, the phase-difference is nearly zero, suggesting a co-movement between the approval rate and consumer sentiment. Subsequently, after 2014, the phase-difference falls between –π/2 and 0, implying that during this period, consumer sentiment leads changes in the approval rating. Furthermore, when examining the estimated coefficient, we observe the same pattern of instability. The estimated coefficient between 2011 and 2012 is lower than the coefficient observed between 2014 and 2016. Figure 11: Wavelet Analysis: Obama’s Approval, Unemployment Rate, Inflation Rate, and Consumer Sentiment. Note: Top: Relation between the Approval rate and the Unemployment rate after controlling for Inflation rate and Consumer Sentiment. Middle: Relation between the Approval rate and the Inflation rate after controlling for Unemployment rate and Consumer Sentiment. Bottom: Relation between the Approval rate and the Consumer Sentiment after controlling for Unemployment and Inflation rates. Charts a.i, b.i and c.i: Partial Wavelet Coherency between the Approval rate and one economic variable after controlling for the others. Diagrams a.ii, b.ii and c.ii: Partial PhaseDifference between the Approval rate and one economic variable after controlling for the others in two frequency bands. Diagrams a.iii, b.iii and c.iii: Partial Wavelet Gain between the Approval rate and one economic variable after controlling for the others in two frequency bands. 56 The wavelet approach sheds, therefore, light on the OLS results. The long-run relation found in the static equation (in which the consumer sentiment’s coefficient was not statistically significant) corresponds to the results that we found for lower frequencies using wavelets. The short-run relation that we found for the consumer sentiment in the dynamical equation corresponds to the result we found at high frequencies. Note, however, that the wavelet results also tell us that the relation between the consumer sentiment and the approval rate is not stable, suggesting that it was particularly strong between 2011 and 2012. In fact, if one reestimates Model 4 of Table 9 after removing the observations between the end of 2011 and 2012, the coefficient is no longer significant at 5%, and it is only marginally significant at 10% (p-value of 0.095). This is the type of information that is almost immediate to obtain with the continuous wavelet transform, which would be hard to obtain with the more traditional techniques. 4.6. Conclusion One of the cornerstones of the literature on government approval is the idea that the state of the economy affects people’s judgment of politicians (Lewis-Beck and Stegmaier, 2013). However, the estimation of the impact of the economy on political popularity has not yielded consistent statistical stability. As a result, several studies have presented divergent findings. This study seeks to contribute to the understanding of this issue by exploring a novel econometric tool. Wavelet analysis allows us to explore nuances that traditional econometric analysis does not. As a complement, wavelet tools can contribute to overcoming some estimation problems, circumventing stationarity and nonlinear issues, and adding new layers of information, allowing the study of the dynamics between two or more variables in the time-frequency domain. In particular, to the extent that the relationship between the economy and political popularity is dynamic and may evolve over time, wavelet analysis is appropriate to distinguish long-term and stable relationships from transient and unstable ones. Moreover, using the wavelet gain in the time-frequency domain (Aguiar-Conraria et al. , 2023), we can even estimate coefficients of multivariate wavelet analysis, an application that has so far remained unexplored in Political Science. In this study, we illustrate this methodological approach to the complex relationship between the economy and the popularity of politicians by examining the approval function for President Barack Obama in a period spanning from January 2009 to December 2016, a period marked by considerable economic instability and uncertainty. Using a traditional econometric approach, we find a negative impact of unemployment and inflation rates in long-run models, while consumer sentiment is only statistically 57 significant when we run dynamic models that capture short-run behavior. Subsequently, we complement these results by running the wavelets tools described in Chapter 4.2. The partial wavelet analysis confirms these results and reinforces the idea that the relationship between economic variables and approval rating is negative and dynamic. The unemployment and inflation rates exhibit a long-term relation with Obama’s approval rating, which is evident throughout his presidency and at lower frequencies. However, the wavelet analysis adds another layer of relevant information. We know that unemployment and inflation have a long-run impact on popularity, but with the phase-difference diagrams, we also know that the time of impact is different. Although both the unemployment and inflation rates lead the approval rate, the synchronization is different. The partial phase-difference diagrams show that the approval rate reacted more quickly to unemployment rate movements than it did to movements of the inflation rate. Furthermore, our analysis also reveals a positive and synchronized co-movement between consumer sentiment and Obama’s approval rating, particularly during the period spanning 2011 and 2012 and at higher frequencies. This transient and unstable relationship between consumer sentiment and Obama’s approval is not delivered by the traditional econometric models. Moreover, the result for consumer sentiment coincides with significant events such as the United States debt-ceiling crisis and the subsequent Federal Government Shutdown, which are responsible for a significant increase in consumers’ worriedness. In conclusion, wavelet analysis is useful for complementing econometric analysis in social sciences, particularly in political science. While traditional econometric tools are important in analyzing and quantifying relationships within a set of variables, they do not allow a time-frequency analysis as the wavelet approach does. 58 5. Political Connections in Portugal 5.1. Introduction Political connections represent an omnipresent aspect of governance and economic systems (Faccio, 2006). At their core, these connections highlight the complex network of relationships between individuals, businesses, and governmental entities, wielding significant influence over resource allocation, policy formulation, and the overall socio-economic landscape. 29 Within the domain of political economy, the study of political connections emerges as a crucial point for understanding the dynamics of power, wealth distribution, and institutional behavior. Within the debate on political connections, there is evidence that this relationship between politics and business has a pervasive impact on economic outcomes (Schoenherr, 2019; Khwaja and Mian, 2005). Regardless of the political regime or the country’s economic and social development, the ability to forge alliances with political elites can confer substantial advantages to firms, ranging from preferential access to public procurement or financial credit, regulatory exemptions to enhanced market influence, and protection from competition. It also confers substantial advantages to politicians, such as an increase in political influence and electoral success (Bertrand et al. , 2018). Such connections are often associated with rent-seeking behavior, whereby individuals or entities seek to extract economic rents through leveraging their political affiliations. Moreover, the opacity surrounding these relationships can undermine transparency, accountability, and the rule of law, fostering environments conducive to corruption and cronyism (Ramalho, 2007; Shleifer and Vishny, 1994). Political connections are not exclusive to developing countries. Even in developed countries with low levels of perceived corruption, the linkage between firms and politicians is a reality (Amore and Bennedsen, 2013). Portugal is no exception, and it is easy to find former politicians as members of the board of directors. Moreover, in a set of surveys conducted by the European Union to businesspeople 29 The literature has also found connections between politicians and other social groups, such as religious leaders (Troncone and Valli, 2024). 59 from all over Europe, most of the Portuguese businesspeople surveyed consistently confirmed that having political connections is the only way to succeed in business (European Commission, 2014, 2015, 2024a). As the literature highlights, this harmful impact on firms and, overall, the country’s economy creates an incentive to study the Portuguese case. Therefore, this Chapter studies the existence and effects of political connections in Portugal. We aim to analyze the characteristics that distinguish connected firms from non-connected firms and whether political connections affect these firms financially. For that purpose, we construct a new database using a sample of publicly traded Portuguese companies. This new database encompasses three types of data: 1) data on the members of the board of directors of the Portuguese publicly traded firms between 2002 and 2022; 2) data on all Presidents of the Republic, members of the government, and members of the parliament since 1974; and, finally, 3) annually data with financial and corporate information between 2002 and 2022. Our findings highlight that firm size, measured by total assets, emerges as a significant factor influencing the presence and the number of politicians on corporate boards. Furthermore, firms in regulated sectors and headquartered in the Lisbon Metropolitan Area show a higher probability of political linkages. We also take into consideration the political ideology, identifying whether the boards are right or left-wing. The results suggest that size explains the predominance of right-wing politically connected members, whereas left-wing connections are more likely in firms headquartered in the Lisbon region. Furthermore, our regression analysis examines the impact of political connections on various financial and market performance indicators. The results indicate that political connections do not significantly influence these financial and market performance measures. This conclusion is robust for different computations of our political connection variable. The main contribution of this Chapter is to enhance a fair debate about the complex relationship between business and politics in Portugal. This relationship has endured since the XIX century and continues to influence society and firms today (Costa et al. , 2010). It is important to understand its existence and societal corporate impacts. Literature has identified potential problems originating from the existence of this kind of relationship, and surveys conducted by the European Union demonstrate it. Portuguese businesspeople have considered political connections as a vehicle that enhances corruption and can hamper business competition (European Commission, 2024a). Therefore, as the literature has highlighted the negative impact of corruption on economic growth (Gründler and Potrafke, 2019), it seems crucial to deepen the study of the existence of political connections in Portuguese firms and to analyze whether they are associated with rent activities and corruption. 60 The chapter is structured as follows. Chapter 5.2 presents a brief review of the literature on political connections, and Chapter 5.3 contextualizes the Portuguese political scenario since the XIX century together with its direct or indirect participation on business. Concerning the empirical part, Chapter 5.4 describes the data collected and how the dataset was constructed, Chapter 4.5 discusses the empirical strategy and the hypothesis formulated, and Chapter 4.6 presents the results. Finally, Chapter 4.7 concludes. 5.2. Literature Review Connections between politicians and firms are a global phenomenon, not confined to poor or developing countries, nor limited to less transparent or autocratic regimes (Faccio, 2006; Faccio, 2010). This is evidenced by the presence of politicians in important positions within both national and multinational companies and shareholders or corporate board members with familial or social ties to politically motivated individuals. Moreover, directors of companies with no political background may run for political office or receive appointments to public offices. When analyzing political connectedness, it is important to consider the institutional and business framework of each country. Certain nations are more susceptible to such connections, which vary in nature and intensity. Generally, countries with a higher incidence of political-business entanglement tend to exhibit greater corruption, impose restrictions on foreign investment abroad by their citizens, and have more transparent systems (Faccio, 2006). 30 Anecdotal evidence suggests a greater prevalence of politically connected firms in developing countries, where corruption levels are high and institutions are weak. Nevertheless, such connections are also found in developed countries with strong institutions and low levels of corruption (Amore and Bennedsen, 2013). The relationship between business and politics is an old story. Firms try to influence political outcomes to gain, for instance, competitive advantages, while politicians support firms for private or electoral gains. Ferguson and Voth (2008) assessed the value of German companies associated with the National Socialist Party (Nazi Party) following Hitler’s appointment as chancellor. They found that firms that supported the German dictator politically and financially enjoyed higher stock returns, regardless of their industry. Similarly, Fascist Italy also granted high rents to certain firms (Faccio and McConnell, 2023), with the fall of the regime causing these firms to underperformance compared to their peers, 30 Transparent systems allow an easy identification of the existing political connections. Transparency can also promote some tolerance regarding connections because any misconduct “is more likely to be punished.” 67 By merging these two datasets, we identified 100 politicians, of which only four are women. To minimize the risk of mismatches, we used various merging strategies and confirmed potential connections using diverse sources, such as online newspapers or directors’ CVs. Overall, we found 144 instances of politicians holding company directorships. Situations where a director assumed a position in a company before their political career are not considered connections. Finally, we collected data on firms’ financial situation. Following Faccio (2010), we collected data on Total Assets (thousands of euros) ( Assets ), Market Capitalization (thousands of euros) ( MCap ), Returns on Equity (%) ( ROE ), Total Debt over Market Capitalization ( Leverage ), Book-to-Market, Market Share, Productivity, and Annual Stock Returns ( Returns ). 39 Market Share is calculated as the proportion of each company’s market capitalization relative to the total market capitalization of all companies in the same industry by year. Finally, Productivity is computed by applying a standard Cobb-Douglas production function, where we regressed the natural logarithm of revenues (output) against the natural logarithms of total assets (capital) and employees (labor), using OLS estimates. We then used the estimated residuals as a measure of multifactor productivity. Industry sectors were identified using the Industry Classification Benchmark ( ICB ). All data was obtained from the Thomson Reuters database. Within the literature on political connections, the topic of privatization is particularly pertinent (Boubakri et al. , 2008). A significant number of post-privatized firms maintain their political connections, and there is a positive correlation with the level of residual state ownership. Therefore, we constructed two additional variables: a dummy variable codded as one for all publicly traded Portuguese companies that underwent privatization or re-privatization ( Privatized ), 40 and another dummy variable codded as one for year/company combinations in which the government held an ownership or shareholder position, including minority stakes ( State Ownership ). To account for political ideology, temporal changes in political connections, and their geographical distribution, we incorporated additional controls. Dummy variables were constructed for prime ministers ( Durao&Santana , Socrates , and PCoelho ), for right-wing governments ( Right ), and for the period during and following the financial intervention and bailout by the Troika group ( Troika ). 41 The 2011 Portuguese Debt Crisis and subsequent intervention by the Troika had a positive impact on transparency, media scrutiny, and regulation of publicly traded companies. 39 All monetized variables were deflated using the Consumer Price Index (CPI), with 2012 as the base year. The CPI data was sourced by INE. 40 These firms were government-owned firms. 41 Following the Global Financial Crisis in 2007 and the beginning of the European Debt Crisis, Prime Minister José Sócrates called for financial support in April 2011. A few days later, a group of negotiators from the European Central Bank, the European Commission, and the International Monetary Fund, known as the Troika , went to Lisbon to negotiate a bailout package. In this package, Portugal committed to equilibrating the public finances, reducing the debt, and privatizing some public companies in exchange for financial aid. 68 Additionally, the Troika mandated the privatization of several companies, reducing the government's presence in various economic sectors. Therefore, the variable Troika is codded as one for the period after 2011 and zero otherwise. For geographic localization, we collected the headquarters address of each publicly traded company and created a dummy variable ( Lisbon ), which equals one if the headquarter is in a municipality of the Lisbon Metropolitan Area and zero otherwise. To control for sector fixed effects, we also created dummies for the following sectors: Consumer staples, energy, financial, health care, industrials, real estate, technology, telecommunications, and utilities. Finally, we created a dummy variable ( Regulated ) codded one for firms in sectors subject to regulatory oversight: energy, financials, telecommunications, and utilities. Table 10 presents the descriptive statistics of the above-mentioned variables. 42 To avoid outliers, we excluded the 0.5th and 99.5th percentiles from Total Assets , ROE , Leverage, B ook-to-Market , and Returns . Regarding corporate finance, the average company has total assets valued at approximately 5,700 million euros despite a considerably high standard deviation. Furthermore, the average company exhibits a return on equity close to 5.6% and a market capitalization near 1,400 million euros. Additionally, Table 11 reports differences in the means of these corporate variables between politically connected firms and non-connected. Regarding firm size, measured by the total assets or the market capitalization, both in logarithms, connected firms are significantly larger. Moreover, connected firms are also more profitable, have relatively higher productivity levels, and are less leveraged. Table 11 also shows that market share is higher for connected firms, but there is no statistical difference in returns between these two groups. Finally, book-to-market is lower for the connected firms, suggesting that these firms are overvalued by investors. 42 Moreover, Table A.4 shows the correlations between all variables described and used in the empirical analysis. 69 Table 10: Descriptive statistics. Observations Mean Std-Dev Min Max Board (N) 907 9.49 5.62 2.00 31.00 Connection 907 0.42 0.49 0.00 1.00 Connection Right 907 0.22 0.42 0.00 1.00 Connection Left 907 0.09 0.28 0.00 1.00 Connection (N) 907 0.85 1.40 0.00 9.00 Total Assets 899 6016451.56 15400121.37 2826.98 94466240.00 Market Cap 902 1469821.35 3041676.12 0.00 20006490.00 ROE 841 5.59 23.40 -171.50 86.54 Leverage 897 3.38 6.03 0.00 54.16 Book to Market 893 1.46 3.34 -18.46 36.50 Market Share 902 23.61 29.06 0.00 100.00 Productivity 762 -0.00 0.70 -2.82 1.97 Returns 886 4.12 44.84 -77.06 299.27 Privatized 907 0.21 0.41 0.00 1.00 State Ownership 907 0.09 0.29 0.00 1.00 Right 907 0.34 0.47 0.00 1.00 Lisbon 907 0.63 0.48 0.00 1.00 Troika 907 0.51 0.50 0.00 1.00 Regulated 907 0.28 0.45 0.00 1.00 Basic Materials 907 0.13 0.34 0.00 1.00 Consumer Discretionary 907 0.18 0.38 0.00 1.00 Consumer Staples 907 0.07 0.25 0.00 1.00 Energy 907 0.02 0.14 0.00 1.00 Financials 907 0.09 0.29 0.00 1.00 Health Care 907 0.02 0.14 0.00 1.00 Industrials 907 0.24 0.43 0.00 1.00 Real Estate 907 0.02 0.12 0.00 1.00 Technology 907 0.07 0.25 0.00 1.00 Telecommunications 907 0.11 0.31 0.00 1.00 Utilities 907 0.06 0.23 0.00 1.00 On average, boards of directors consist of 9.5 members, with a range from 2 to 31 members (Table 10). Moreover, the average number of connections per board is 0.85, with some boards having up to 9 politically connected members ( Connection (N) ). Figure 12 illustrates the evolution of the average number of connections from 2002 to 2022. There was an upward trend from 2002 to 2011, with the average number of politically connected members per board increasing from 0.69 in 2002 to 1.07 in 2011. Subsequently, after 2011, there was a significant decrease in the average number of politically connected members, plummeting to 0.31. This decline was spurred by a reduction in politically connected members 70 within boards despite no parallel reduction in board size. Figure 1313 highlights this declining trend, particularly evident from around 2006 onward and notably pronounced from 2016 onwards. Table 11: T-test on the differences in the means of financial indicators between politically connected and non-connected firms. Connected Non-Connected Diff Log TA 14.76 12.61 2.15*** Market Cap 13.48 11.06 2.42*** Roe 7.10 4.41 2.69* Leverage 2.94 3.71 -0.78* Market Share 33.13 16.59 16.54*** Returns 3.24 4.78 -1.54 Productivity 0.06 -0.06 0.12** Book to Market 1.07 1.76 -0.69*** Note: Statistical significance: ***p<.01; **p<.05; *p<.1. Figure 12: Average number of board members politically connected by year (2002-2022). 71 Certain industrial sectors, especially those involving non-tradable goods (Domadenik et al. , 2016) and regulated industries (Boubakri et al. , 2008), seem to be more prone to political connections. Table 12 12 shows that companies within the Energy, Utilities, Financials, and Telecommunication sectors typically have boards of directors with more than ten members and a higher proportion of former politicians. Utilities have the highest average of politically connected members (3.14), and approximately 16.2% of Utility board members are former politicians, in contrast to 7% across all sectors. Doing an analysis between non-regulated and regulated sectors, Figure 14 shows that in all years, the regulated sector highlights higher percentages of politically connected members than the non-regulated sector. Moreover, the percentage of politically connected members from the non-regulated sector has exhibited a declining trend since 2005, overlapping the declining trend observed in Figure 13. However, the percentage from the regulated sector maintained above 10% until 2019, when it started to decrease. In 2022, the percentage of politically connected members from the regulated sector was less than 5%. Figure 13: Average number of board members, average number of board members politically connected, and percentage of board members politically connected by year (2002-2022). 72 Table 12: Average number of board members and political connections by industry. Industry Average number of board members Political Connections Average number (%) Basic Materials 7.20 0.26 3.34 Consumer Discretionary 6.63 0.70 8.92 Consumer Staples 7.88 0.38 3.99 Energy 18.53 2.88 15.39 Financials 14.20 1.76 10.97 Health Care 7.37 0.07 0.44 Industrials 7.67 0.48 4.55 Real Estate 3.57 0.00 0.00 Technology 8.15 0.46 5.67 Telecommunications 10.73 0.83 7.46 Utilities 17.40 3.14 16.18 Overall 9.94 0.99 6.99 Figure 14: Percentage of politically connected members on boards in regulated and non-regulated firms by year. 73 5.5. Empirical Strategy In this Subchapter, we describe the empirical methodology used to examine the relationship between political connections and Portuguese publicly traded firms, as discussed in the introduction. Our approach aims to analyze how a set of characteristics, such as firm size, localization, industry, and regulation, influence the likelihood of a firm having politically connected board members. Additionally, we assess the impact of political connections on key financial indicators. 5.5.1. Firm Characteristics and Political Connections To study the characteristics that could distinguish firms with and without political connections, we employ probit regression models to estimate the likelihood of a firm being politically connected. The dependent variable is a dummy variable that equals one if a firm’s board has at least one former minister, state secretary, or member of parliament as a director in a given year and zero otherwise (see variable Connection in Table 10). This variable identifies the presence of political links within the board of each company and year. Probit models are particularly suited for this kind of analysis, as they allow us to model the probability of an event occurring while accounting for the fact that the dependent variable is binary. In this particular case, the probit model assumes that the relationship between the covariates (firm size, localization, industry, etc.) and the probability of being politically connected is nonlinear and follows a normal distribution. The general model to be estimated can be represented by the equation (1): 𝑌𝑖,𝑡 =𝛽0+𝛾.𝑿𝑖,𝑡 +𝜇𝑖+𝜆𝑡+𝜀𝑖,𝑡, (1) where 𝑌𝑖,𝑡 is the dependent variable, which is our dummy variable that captures the presence of political connections for each Portuguese publicly traded firm in year t. 𝑿𝑖,𝑡 is a vector of firm’s characteristic, 𝜆𝑡 represents time-fixed effects and 𝜇𝑖 industry fixed-effects. 43 𝛽0 is a constant, 𝛾 a vector of parameters, and 𝜀𝑖,𝑡 the error term. As for the firm’s characteristics, we included the logarithm of total assets to measure the firm’s size, a dummy variable to control for the existence of a past privatization or re43 Estimating nonlinear models like Probit with individual fixed effects can lead to a phenomenon known as the incidental parameters problem when T is not sufficiently large. To avoid this issue, we used industry-fixed effects instead. For robustness, we also estimated logit models with industry-fixed effects, and the results were similar. 74 privatization process or if the firm has a public stake, and a dummy variable that identifies those firms that have headquarters in the Lisbon region. Furthermore, we also included dummy variables for each Portuguese government and a dummy for firms operating in regulated industries. The literature has identified the characteristics that allow the identification of politically connected firms. As explained in Chapter 5.2, size, privatization or state presence, localization, or regulation are important variables to study political connectedness. Therefore, we anticipate that larger firms or firms that are privatized or have a public stake have a higher probability of being politically connected. Moreover, firms whose headquarters are in the Lisbon region, close to the political decision-making center, are expected to have more former politicians on their boards. Finally, firms from regulated sectors (energy, financials, telecommunications, and utilities) may exhibit more politically connected boards because of the possibility of influencing the decisions regarding industry regulation. 5.5.2. Impact on Corporate Variables The next step is to determine whether the presence of politically connected board members influences key financial outcomes such as profitability, leverage, and market share. Here, we explore the methodology used to examine the effect of political connections on the financial performance of Portuguese publicly traded firms. We run ordinary least squares (OLS) models to assess the relationship between political connections and financial variables while controlling for the firm’s characteristics. Therefore, our main independent variable of interest is the dummy variable Connection . Thus, the estimated equation is the following: 𝒀𝒊,𝒕 =𝛽0+𝛽1.𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑖𝑜𝑛+𝛾.𝑿𝑖,𝑡 +𝜇𝑖+𝜆𝑡+𝜀𝑖,𝑡, (2) where 𝒀𝒊,𝒕 is a vector of dependent variables, and 𝑿𝑖,𝑡 a vector of firm’s characteristics. The variable Connection is the most important independent variable and, therefore, if statistically significant, the estimate 𝛽1 for each model captures differences in firms’ financial performance depending on the existence of political connections. To control for firm’s characteristics, we included its size and localization. 𝛽0 is a constant, 𝛾 a vector of parameters, and 𝜀𝑖,𝑡 the error term. Finally, 𝜆𝑡 represents time-fixed effects and 𝜇𝑖 industry fixed-effects. As dependent variables, we chose a set of financial variables commonly used in the literature: returns on equity (ROE), leverage, book-to-market, market share, productivity, and stock returns. Based 75 on the literature, we expect that political connections may have a mixed impact on these financial outcomes. Previous studies (Faccio, 2010; Desai and Olofsgard, 2011) suggest that politically connected firms often underperform in terms of profitability (ROE) and productivity due to inefficiencies arising from rent-seeking behavior and resource misallocation. Moreover, these firms may be more leveraged, relying on easier access to credit (Bussolo et al. , 2022). On the other hand, political connections could lead to higher market shares or higher book-to-market (Faccio, 2010). Finally, these firms are also expected to enjoy higher stock returns. However, the literature also highlights that these firms suffer more from political cycles and economic decline, and the results can be mixed. 5.6. Results Table 13 presents the marginal effects of the models estimated by Probit regression models. These models aim to estimate the probability of a firm being politically connected while controlling for various firm’s characteristics. In Model 1, the probability of a firm having political connections increases with its size, measured by the natural logarithm of total assets. Specifically, the marginal effect of a one percent increase in total assets on the conditional probability of having at least one politician is 0.14 percentage points. The presence of the state on the shareholder structure of the firm also has a positive marginal effect on the probability of being connected, with a coefficient of 0.196. In Model 2, we replaced the variable State Ownership with the variable Privatized . It is noteworthy that almost all firms that were privatized or reprivatized went through several privatization phases, which means that privatized firms had the state as shareholders during these periods. Additionally, some firms continue to have the state as a shareholder. Thus, the variable Privatized serves as an alternative proxy to control for the state’s presence and the potential influence on the firm boards. The coefficient is almost the same in magnitude and statistically significant, slightly stronger than the State Ownership ’s. Geographic proximity to the political decision center was also considered. Portugal is considerably centralized around its capital, Lisbon, where all political institutions and regulators’ headquarters are located. 44 Model 3 introduces a variable to control for the localization of the firm’s headquarters ( Lisbon ). The coefficient is positively signed and statistically significant, as we expected. Therefore, firms headquartered near the political decision-making center have a higher likelihood of political connections, with a marginal effect of 0.193. Firms’ size continues to be an important characteristic, whereas the State 44 The exception is the Portuguese Health Regulatory Authority ( ERS ), located in Oporto. There were some attempts to move the National Authority of Medicines and Health Products ( Infarmed ) to Oporto and the Portuguese Constitutional Court to Coimbra, but unsuccessfully. 76 Ownership coefficient turns out to be statistically not significant. As a robustness check, we also regressed Model 3 with the variable Privatized instead of State Ownership (not shown in the Table), and the results were similar. Thus, the Lisbon localization seems to be a better predictor of the probability of a firm having members of the board politically connected. This result is corroborated by the fact that only one privatized firm was headquartered outside of the Lisbon Metropolitan Region. Therefore, the variable Lisbon is also likely to capture the effect of the presence of the state as a shareholder. Consequently, the variables controlling for the presence of the State on firms’ ownership were dropped in the following models of Table 13. To avoid losing a considerable number of degrees of freedom, year-fixed effects were replaced with three dummy variables (column 4), each signaling the tenure of a prime minister, ranging from Durão Barroso/Santana Lopes to Passos Coelho (see Table A.2 in appendix). 45 This adjustment yielded consistent findings for firm size and localization. The prime ministers’ coefficients are all positive and statistically significant, exhibiting a decreasing trend over time. This suggests that the probability of a firm having politically connected directors decreased over time, confirming the trend observed in Figure 13. Additionally, in Model 5, we replaced the variables for prime ministers with a dummy variable that captures the troika period. As expected, the coefficient is negative and statistically significant. 46 Regarding the industry, columns 1 to 6 present the estimated coefficients for each sector. 47 Results indicate that companies from the sectors of Consumer Discretionary, Technology, Telecommunications, and Utilities are more likely to have boards with political connections. As some industries are subjected to market regulation, in Model 6, instead of using industry-fixed effects, we used a variable that captures the dependency of a specific industry on regulatory procedures (R egulated ). There is evidence that being a firm in a regulated sector increases the probability of having a political connection, as indicated by a marginal effect of 0.199, which is statistically significant. Overall, size, headquarters’ localization, and being in a regulated sector are characteristics that augment the probability of political linkages. These results are consistent with the existing literature, as evidenced by Faccio (2010) and Boubakri et al. (2008). We also tried the logarithmic transformation of 45 The base scenario is the governments of PS from 2016 to 2022. We did not create two separate dummy variables for Durão Barroso and Santana Lopes since Santana’s government had a short duration, lasting less than a year, and was from the same party as its predecessor. 46 We also run a model substituting the dummy variable Troika with a categorical variable, where the period before 2011 is codded as 1, the period between 2011 and 2014 is codded as 2, and the period after 2014 is codded as 3. Considering the period pre-troika as the base category, the result shows that the probability of having political connections is only statistically lower after 2014. Results are available upon request. 47 The Energy and the Real Estate sectors have only one company. Therefore, their coefficients do not appear in Table 13. 83 5.7. Conclusion In conclusion, this study delved into the existence and impact of political connections in Portuguese publicly traded companies. By analyzing the composition of boards of directors and their former political positions, we aimed to uncover the characteristics that distinguish connected and non-connected firms and how political connections might influence a firm’s financial outcomes. Our analysis spanned two decades, focusing on firms listed on the Euronext Lisbon stock exchange. This research contributes to the broader understanding of the interaction between politics and business, shedding light on the prevalence and implications of political ties within corporate governance in Portugal. To conduct this analysis, we constructed three new primary datasets. The first dataset comprised the names of all board members from Portuguese publicly traded companies since 2002. The second dataset included the names of all politicians who held office since the 1974 coup d’état until March 2022. This list encompassed presidents of the republic, members of interim and constitutional governments, and members of Parliament. Additionally, we gathered extensive data on firms’ financial conditions, including variables such as Total Assets, Market Capitalization, Returns on Equity, Leverage, Book-toMarket, Market Share, and Productivity. These datasets enabled a thorough examination of the potential influence of political connections on firm performance. Our findings reveal several key insights. First, larger firms, firms in regulated sectors, and firms headquartered in the Lisbon Metropolitan Area show a higher probability of political linkages. However, the number of politically connected directors tends to decrease over time, and differences in that number are not significantly influenced by state regulation in the sector. Interestingly, while government ideology plays a role, firm size, and headquarters’ location are more critical determinants of political connections. Boards with a predominance of right-wing politically connected members are typically larger, whereas leftwing politically connected members are more prevalent in firms headquartered in the Lisbon Metropolitan Area. Notably, our regression analysis indicates that political connections do not significantly impact key financial and market performance indicators. Thus, while political connections exist in Portuguese publicly traded firms, although they have lost strength over time, they do not appear to confer a significant financial advantage or disadvantage. This last result aligns with Keefe (2019), highlighting that heterogeneous institutional contexts and specific mechanisms may explain the inconsistent results found in the literature. Moreover, this Chapter looks at political connections between publicly traded firms, overseen by the stock market regulator and under scrutiny by the media, with “high-profile political connections.” For Akcigit et 84 al. (2023), the links with local politicians “[…] are much more pervasive and can have broader consequences for the overall economy.” This Chapter provides valuable insights into the relationship between politics and firms, particularly in the context of a developed country like Portugal. However, several limitations should be acknowledged. First, our dataset is restricted to publicly traded companies, which may not fully capture the dynamics of political connections in privately held firms. Moreover, the Portuguese stock market consists of a relatively small number of companies when compared to other European stock markets. Second, our measure of political connections is based on the explicit presence of former politicians on boards, potentially ignoring other forms of political influence that may also play a significant role and that have been discussed in the literature. For future research, expanding the dataset to include a broader range of firms and additional years could provide a more comprehensive picture of political connections in Portugal. Furthermore, exploring other channels through which political influence operates could deepen our understanding of how these ties impact firms’ performance. 85 6. Conclusion This dissertation delves into some topics studied in the field of Political Economy. The first three essays discuss and present potential solutions for the unstable and sometimes contradictory impact of the economy on political popularity. The last essay analyzes political connections, a topic studied by different economic branches, including the political economy. Chapter 2 presents a methodological alternative that uses different economic measures. The studies of popularity functions use accurate economic data to measure the impact of the economy on popularity. This essay distinguishes between the “objective” and “mediate” economies. The “objective” economy is the data that reflects the true developments of economic growth, whereas the “mediate” economy reflects the economic information available to the public by the media. We expect voters to use the economic information portrayed by the media rather than the most accurate economic information. The essay also introduces projected economic growth for the current and the following year, which is information made public by the media. The essay reveals that projections for GDP growth, rather than actual or retrospective economic performance, are the strongest predictor of prime ministerial approval in Portugal, using quarterly data from 2001 to 2018. This finding emphasizes that voters rely on mediated economic information, particularly regarding future expectations when assessing government performance. The limitations of this study include its focus on Portugal, suggesting a need for cross-national research to confirm the applicability of these findings. Additionally, future research should examine individual-level data to understand better how economic forecasts and media framing shape public perceptions and political judgments. This could enhance forecasting models in the political economy by incorporating real-time projections, potentially improving the accuracy of election forecasts. Chapter 3 presents another methodological alternative by considering a specific political system: the semi-presidentialism. It investigates whether the dynamics observed in other semi-presidential contexts—where cohabitation often shifts economic accountability between executive leaders—apply in Portugal, where the president is the head of state and the prime-minister is the head of the executive 86 branch. Therefore, it assesses whether the relationship between both political entities, under cohabitation or unified government, changes the responsibility for economic development. Additionally, it seeks to identify potential asymmetries in how economic evaluations affect prime ministers versus presidents. The essay finds that in Portugal’s semi-presidential system, between 1986 and 2018, prime ministers, not presidents, are primarily held accountable for economic performance, whether under cohabitation or unified government. A notable exception arises during conflictual cohabitations in second presidential terms, where presidential opposition seems to blur economic accountability, making public approval for both leaders less sensitive to economic conditions. We also observe an asymmetry: economic downturns impact prime ministerial approval more than presidential, reflecting differing partisan expectations. This insight into Portugal’s V-P dynamics suggests that in similar semi-presidential systems, economic accountability may similarly vary across leaders, especially under cohabitation conflicts. Future research should explore these patterns in comparable political contexts. Chapter 4 aims to discuss a potential solution for the unstable results of the impact of the economy on popularity by utilizing a new set of tools. This essay explores how wavelet analysis can help political economic researchers circumvent some traditional econometric limitations. By applying multivariate wavelet analysis, which allows a time-frequency domain analysis, this essay seeks to capture both stable long-term patterns and transient fluctuations in political approval. To do so, we focused on Barack Obama’s presidency, the last two-term US president (2009-2017), which was associated with the economic recovery after the 2009 Great Depression. The results show a heterogeneous relationship between economic variables and Obama’s approval rating. Traditional models reveal that unemployment and inflation have a long-term negative effect on Obama’s approval, with consumer sentiment affecting approval only in short-run dynamic models. The wavelet analysis confirms these findings, adding detail by showing that unemployment influences approval ratings more quickly than inflation despite both leading approval rate changes over the long run. Additionally, wavelet analysis reveals a positive, yet transient, connection between consumer sentiment and approval in 2011-2012, corresponding to the first Obama debt-ceiling crisis event. While wavelet analysis offers valuable insights into the timing and frequency of these relationships, future research could extend this method to examine other political contexts in the political economy and political science fields. Finally, Chapter 5 aims to investigate the political connections in publicly traded Portuguese companies by examining the political backgrounds of board members and their potential influence on firm performance. Focusing on public trade firms listed on the Euronext Lisbon stock exchange over the 87 period 2002-2022, we constructed three unique datasets detailing board members, Portuguese politicians, and firms’ financial outcomes. Our objective is to provide insights into the characteristics of firms that have political connections and how political ties may influence financial outcomes. Our findings reveal that larger firms, those based in Lisbon, and those in regulated sectors are more likely to have political connections, though these connections have weakened over time and have a limited impact on key financial outcomes. Right-leaning politically connected members are more common on larger boards while left-leaning connections are more prevalent in firms headquartered in Lisbon. Importantly, political connections do not appear to significantly impact financial performance. Limitations of this study include the narrow focus on publicly traded firms and the specific measurement of political connections, which may miss other forms of political influence. Future research could expand this analysis to include privately held firms and alternative political influence channels to build a more comprehensive understanding of political ties in corporate performance and in the overall economy. 88 Bibliography Acemoglu, D., Hassan, T. A., & Tahoun, A. (2017). The power of the street: Evidence from egypt’s arab spring. The Review of Financial Studies , 31 (1), 1-42. Acemoglu, D., Johnson, S., Kermani, A., Kwak, J., & Mitton, T. (2016). The value of connections in turbulent times: Evidence from the United States. Journal of Financial Economics , 121 (2), 368391. Acevedo, D. N., Fogleman, C., & Daniel Ura, J. (2017). Peasants and bankers: Education, consumer sentiment, and presidential approval. Presidential Studies Quarterly , 47 (2), 230-244. Agrawal, A., & Knoeber, C. R. (2001). Do some outside directors play a political role?. The Journal of Law and Economics , 44 (1), 179-198. Aguiar‐Conraria, L., & Soares, M. J. (2014). The continuous wavelet transform: Moving beyond uni‐and bivariate analysis. Journal of economic surveys , 28 (2), 344-375. Aguiar‐Conraria, L., Magalhães, P. C., & Soares, M. J. (2012). Cycles in politics: Wavelet analysis of political time series. American Journal of Political Science , 56 (2), 500-518. Aguiar-Conraria, L., Magalhães, P. C., & Soares, M. J. (2013). The nationalization of electoral cycles in the United States: a wavelet analysis. Public Choice , 156 (3-4), 387-408. Aguiar-Conraria, L., Martins, M. M., & Soares, M. J. (2023). The Phillips curve at 65: Time for time and frequency. Journal of Economic Dynamics and Control , 151 , 104620. Akcigit, U., Baslandze, S., & Lotti, F. (2023). Connecting to power: political connections, innovation, and firm dynamics. Econometrica , 91 (2), 529-564. Alt, J. E., Marshall, J., & Lassen, D. D. (2016). Credible sources and sophisticated voters: When does new information induce economic voting?. The Journal of Politics , 78 (2), 327-342. Amore, M. D., & Bennedsen, M. (2013). The value of local political connections in a low-corruption environment. Journal of Financial Economics , 110 (2), 387-402. Anderson, C. J. (2007). The end of economic voting? Contingency dilemmas and the limits of democratic accountability. Annual Review of Political Science , 10 , 271-296. Anderson, C. J., & Tverdova, Y. V. (2003). Corruption, political allegiances, and attitudes toward government in contemporary democracies. American Journal of Political Science , 47 (1), 91-109. Becher, M., & Donnelly, M. (2013). Economic performance, individual evaluations, and the vote: Investigating the causal mechanism. The Journal of Politics , 75 (4), 968-979. Bélanger, É., & Meguid, B. M. (2008). Issue salience, issue ownership, and issue-based vote choice. Electoral Studies , 27 (3), 477-491. Berlemann, M., & Enkelmann, S. (2014). The economic determinants of US presidential approval: A survey. European Journal of Political Economy , 36 , 41-54. Berlemann, M., Enkelmann, S., & Kuhlenkasper, T. (2015). Unraveling the relationship between presidential approval and the economy: A multidimensional semiparametric approach. Journal of Applied Econometrics , 30 (3), 468-486. Bertrand, M., Kramarz, F., Schoar, A., & Thesmar, D. (2018). The cost of political connections. Review of Finance , 22 (3), 849-876. Blau, B. M., Brough, T. J., & Thomas, D. W. (2013). Corporate lobbying, political connections, and the bailout of banks. Journal of Banking & Finance , 37 (8), 3007-3017. Boomgaarden, H. G., Van Spanje, J., Vliegenthart, R., & De Vreese, C. H. (2011). Covering the crisis: Media coverage of the economic crisis and citizens’ economic expectations. Acta Politica , 46 (4), 353-379. Boubakri, N., Cosset, J. C., & Saffar, W. (2008). Political connections of newly privatized firms. Journal of Corporate Finance , 14 (5), 654-673. 89 Boukes, M., Damstra, A., & Vliegenthart, R. (2021). Media effects across time and subject: How news coverage affects two out of four attributes of consumer confidence. Communication Research , 48 (3), 454-476. Boydstun, A. E., Highton, B., & Linn, S. (2018). Assessing the relationship between economic news coverage and mass economic attitudes. Political Research Quarterly , 71 (4), 989-1000. Brender, A., & Drazen, A. (2007). Electoral fiscal policy in new, old, and fragile democracies. Comparative Economic Studies , 49 (3), 446-466. Brody, R. (1991). Assessing the president: The media, elite opinion, and public support . Stanford University Press. Brown, J. R., & Huang, J. (2020). All the president's friends: Political access and firm value. Journal of Financial Economics , 138 (2), 415-431. Bussolo, M., de Nicola, F., Panizza, U., & Varghese, R. (2022). Politically connected firms and privileged access to credit: Evidence from Central and Eastern Europe. European Journal of Political Economy , 71 , 102073. Carlin, R. E., Hartlyn, J., Hellwig, T., Love, G. J.,Martinez-Gallardo, C., & Singer, M. M. (2019). Executive Approval Database 2.0 . http://www.executiveapprovaldata.org Carlin, R. E., Hartlyn, J., Hellwig, T., Love, G. J.,Martinez-Gallardo, C., & Singer, M. M. (2023). Introduction: Instability and government popularity in the twenty-first century. In T. Hellwig, & M. M. Singer (Eds.), Economics and politics revisited: Executive approval and the new calculus of support (pp. 1–31). Oxford University Press. Child, T. B., Massoud, N., Schabus, M., & Zhou, Y. (2021). Surprise election for Trump connections. Journal of Financial Economics , 140 (2), 676-697. Choi, J. P., & Thum, M. (2009). The economics of politically-connected firms. International Tax and Public Finance , 16 , 605-620. Choi, S. W., James, P., Li, Y., & Olson, E. (2016). Presidential approval and macroeconomic conditions: Evidence from a nonlinear model. Applied Economics , 48 (47), 4558-4572. Chzhen, K., Evans, G., & Pickup, M. (2014). When do economic perceptions matter for party approval?. Political Behavior , 36 (2), 291-313. Claessens, S., Feijen, E., & Laeven, L. (2008). Political connections and preferential access to finance: The role of campaign contributions. Journal of Financial Economics , 88 (3), 554-580. Cohen, J. E. (2004). Economic perceptions and executive approval in comparative perspective. Political Behavior , 26 (1), 27-43. Costa, J., Fazenda, L., Honório, C., Louçã, F., & Rosas, F. (2010). Os donos de Portugal: cem anos de poder económico (1910-2010) . Edições Afrontamento. Coulomb, R., & Sangnier, M. (2014). The impact of political majorities on firm value: Do electoral promises or friendship connections matter? Journal of Public Economics , 115 , 158-170. Damstra, A., & Boukes, M. (2021). The economy, the news, and the public: A longitudinal study of the impact of economic news on economic evaluations and expectations. Communication Research , 48 (1), 26-50. Dassonneville, R., & Lewis-Beck, M. S. (2014). Macroeconomics, economic crisis and electoral outcomes: A national European pool. Acta Politica , 49 (4), 372-394. De Boef, S., & Kellstedt, P. M. (2004). The political (and economic) origins of consumer confidence. American Journal of Political Science , 48 (4), 633-649. De Sousa, L., & Moriconi, M. (2013). Why voters do not throw the rascals out?—A conceptual framework for analysing electoral punishment of corruption. Crime, Law and Social Change , 60 , 471-502. Desai, R. M., & Olofsgård, A. (2011). The costs of political influence: Firm-level evidence from developing countries. Quarterly Journal of Political Science , 6 (2), 137-178. 90 Dickerson, B. (2016). Economic perceptions, presidential approval, and causality: The moderating role of the economic context. American Politics Research , 44 (6), 1037-1065. Dinç, I. S. (2005). Politicians and banks: Political influences on government-owned banks in emerging markets. Journal of Financial Economics , 77 (2), 453-479. Domadenik, P., Prašnikar, J., & Svejnar, J. (2016). Political connectedness, corporate governance, and firm performance. Journal of Business Ethics , 139 (2), 411-428. Duch, R. M., & Stevenson, R. T. (2008). The economic vote: How political and economic institutions condition election results . Cambridge University Press. Duch, R. M., Palmer, H. D., & Anderson, C. J. (2000). Heterogeneity in perceptions of national economic conditions. American Journal of Political Science , 635-652. Duverger, M. (1980). A new political system model: Semi‐presidential government. European journal of political research , 8 (2), 165-187. Elgie, R. (1999). The politics of semi-presidentialism . In: R. Elgie (Ed.), Semi-presidentialism in Europe (pp. 1-21) . Oxford University Press. Elgie, R. (2010). Semi-presidentialism, cohabitation and the collapse of electoral democracies, 1990– 2008. Government and Opposition , 45 (1), 29-49. Elgie, R. (2018). Regime types, presidential power, and clarity of economic responsibility. In R. Elgie (Ed.), Political Leadership: A Pragmatic Institutionalist Approach (pp. 85-125) . Palgrave Macmillan. Erikson, R. S., MacKuen, M. B., & Stimson, J. A. (2000). Bankers or peasants revisited: economic expectations and presidential approval. Electoral Studies , 19 (2-3), 295-312. European Commission (2014). Flash Eurobarometer 374: Businesses' attitudes towards corruption in the EU. European Union. https://europa.eu/eurobarometer/surveys/detail/1077. European Commission (2015). Flash Eurobarometer 428: Businesses' attitudes towards corruption in the EU. European Union. https://europa.eu/eurobarometer/surveys/detail/2084. European Commission (2024a). Flash Eurobarometer 543: Businesses' attitudes towards corruption in the EU. European Union. https://europa.eu/eurobarometer/surveys/detail/3180. European Commission (2024b). Special Eurobarometer 548: Citizens' attitudes towards corruption in the EU in 2024. European Union. https://europa.eu/eurobarometer/surveys/detail/3217 Evans, G., & Andersen, R. (2006). The political conditioning of economic perceptions. The Journal of Politics , 68 (1), 194-207. Faccio, M. (2006). Politically connected firms. The American Economic Review , 96 (1), 369-386. Faccio, M. (2010). Differences between politically connected and nonconnected firms: A cross‐country analysis. Financial Management , 39 (3), 905-928. Faccio, M., & Hsu, H. C. (2017). Politically connected private equity and employment. The Journal of Finance , 72 (2), 539-574. Faccio, M., & McConnell, J. J. (2023). Political connections cause resource misallocation: Evidence from the fall of fascism in Italy. ECGI Working Paper Series in Finance , (939). Faccio, M., & Zingales, L. (2022). Political determinants of competition in the mobile telecommunication industry. The Review of Financial Studies , 35 (4), 1983-2018. Faccio, M., Masulis, R. W., & McConnell, J. (2006). Political connections and corporate bailouts. The Journal of Finance , 61 (6), 2597-2635. Feijó, R. G. (2021). "Moderating power" and partisan presidents: Two empirical cases. In R. G. Feijó (Ed.), Presidents in semi-presidential regimes (pp. 129-149). Palgrave Pivot. Ferguson, T., & Voth, H. J. (2008). Betting on Hitler—the value of political connections in Nazi Germany. The Quarterly Journal of Economics , 123 (1), 101-137. Fernandes, F. (2018). Os empresários de Marcello Caetano: Os grandes anos de investimento público e privado, entre 1968-1974, e as guerras empresariais . Casa das Letras. 91 Fisman, R. (2001). Estimating the value of political connections. The American Economic Review , 91 (4), 1095-1102. Fortes, B. G., & Magalhães, P. (2005). As eleições presidenciais em sistemas semipresidenciais: participação eleitoral e punição dos governos. Análise Social , 891-922. Franco, V. S. D., 2020. Semipresidencialismo em Portugal: Os poderes presidenciais na interação com o governo e com a Assembleia da República (1982-2016). Lisbon: Assembleia da República. Freire, A., & Lobo, M. C. (2005). Economics, ideology and vote: Southern Europe, 1985–2000. European Journal of Political Research , 44 (4), 493-518. Freire, A., & Santana-Pereira, J. (2012). Economic voting in Portugal, 2002–2009. Electoral Studies , 31 (3), 506-512. Freire, A., & Santana-Pereira, J. (2019). The President's dilemma: The Portuguese semi-presidential system in times of crisis (2011‐16). International Journal of Iberian Studies , 32 (3), 117-135. Goidel, R. K., & Langley, R. E. (1995). Media coverage of the economy and aggregate economic evaluations: Uncovering evidence of indirect media effects. Political Research Quarterly , 48 (2), 313-328. Goodhart, C. A. E & Bhansali, R. J. (1970). Political Economy. Political Studies , 18 (1), 43-106. Goulart, P., & Veiga, F. J. (2016). Portuguese 2015 legislative elections: How economic voting, the median voter and unemployment led to ‘the times they are a’changin’? Electoral Studies , 43 , 197–200. Gronke, P., & Newman, B. (2003). FDR to Clinton, Mueller to?: A field essay on presidential approval. Political Research Quarterly , 56 (4), 501-512. Grossman, E., & Guinaudeau, I. (2023). Cost of ruling above all else: Explaining presidential popularity in France. In T. Hellwig, & M. M. Singer (Eds.), Economics and politics revisited: Executive approval and the new calculus of support (pp. 80–107). Oxford University Press. Gründler, K., & Potrafke, N. (2019). Corruption and economic growth: New empirical evidence. European Journal of Political Economy , 60 , 101810. Hao, Z., Liu, Y., Zhang, J., & Zhao, X. (2020). Political connection, corporate philanthropy and efficiency: Evidence from China’s anti-corruption campaign. Journal of Comparative Economics , 48 (3), 688708. Harstad, B., & Svensson, J. (2011). Bribes, lobbying, and development. American Political Science Review , 105 (1), 46-63. Hellwig, T., & Marinova, D. M. (2017). Dead ends and new paths in the study of economic voting. In Oxford Research Encyclopedia of Politics . Oxford University Press. Hellwig, T., & Samuels, D. (2008). Electoral accountability and the variety of democratic regimes. British Journal of Political Science, 38 (1), 65-90. Hibbs, D. A. (1977). Political parties and macroeconomic policy. The American Political Science Review , 71 (4), 1467-1487. Jalali, C. (2011). The president is not a passenger. In R. Elgie, S. Moestrup, & Y.-S. Wu (Eds.), Semipresidentialism and democracy (pp. 156-173). Palgrave Macmillan. Johnson, S., & Mitton, T. (2003). Cronyism and capital controls: evidence from Malaysia. Journal of Financial Economics , 67 (2), 351-382. Kayser, M. A. (2014). The elusive economic vote. In L. Leduc & R. Niemi (Eds.), Comparing democracies: Elections and voting in a changing world (Vol. 4, pp.112-132). Thousand Oaks, CA: Sage. Kayser, M. A., & Leininger, A. (2015). Vintage errors: do real-time economic data improve election forecasts?. Research & Politics , 2 (3), 1-11. Keefe, M. O. C. (2019). A theory of political connections and financial outcomes. International Review of Economics & Finance , 61 , 108-127. 92 Khwaja, A. I., & Mian, A. (2005). Do lenders favor politically connected firms? Rent provision in an emerging financial market. The Quarterly Journal of Economics , 120 (4), 1371-1411. Lacy, D., & Christenson, D. P. (2017). Who votes for the future? Information, expectations, and endogeneity in economic voting. Political Behavior , 39 (2), 347-375. Lehrer, N. D. (2018). The value of political connections in a multiparty parliamentary democracy: evidence from the 2015 elections in Israel. European Journal of Political Economy , 53 , 13-58. Lewis‐Beck, M. S. (1997). Who's the chef? Economic voting under a dual executive. European Journal of Political Research , 31 (3), 315-325. Lewis-Beck, M. S. (2004). French election theories and the 2002 results: An introduction. In M. S. LewisBeck (Ed.), The French voter (pp. 1-11). Palgrave Macmillan. Lewis-Beck, M. S., & Nadeau, R. (2004). Dual governance and economic voting: France and the United States. In M. S. Lewis-Beck (Ed.), The French voter (pp. 136-154). Palgrave Macmillan. Lewis-Beck, M. S., & Nadeau, R. (2012). PIGS or not? Economic voting in Southern Europe. Electoral Studies , 31 (3), 472-477. Lewis-Beck, M. S., & Stegmaier, M. (2000). Economic determinants of electoral outcomes. Annual Review of Political Science , 3 (1), 183-219. Lewis-Beck, M. S., & Stegmaier, M. (2013). The VP-function revisited: a survey of the literature on vote and popularity functions after over 40 years. Public Choice , 157 , 367-385. Lijphart, A. (2012). Patterns of democracy: Government forms and performance in thirty-six countries . Yale University Press. Lobo, M. C., & Pannico, R. (2020). Increased economic salience or blurring of responsibility? Economic voting during the Great Recession. Electoral Studies , 65 , 102141. MacKuen, M. B., Erikson, R. S., & Stimson, J. A. (1992). Peasants or bankers? The American electorate and the US economy. American Political Science Review , 86 (3), 597-611. Magalhães, P. C. (2001). As armas dos fracos: O veto político e a litigância constitucional do Presidente da República. In Actas do I Encontro Nacional de Ciência Política . Bizâncio. Magalhães, P. C. (2017). A tale of two elections: information, motivated reasoning, and the economy in the 2011 and 2015 Portuguese elections. Análise Social , 52 (225), 736-758. Magalhães, P. C., & Aguiar-Conraria, L. (2009). Growth, centrism and semi-presidentialism: Forecasting the Portuguese general elections. Electoral Studies , 28 (2), 314-321. Mandler, M., & Scharnagl, M. (2014). Money growth and consumer price inflation in the euro area: A wavelet analysis. Bundesbank Discussion Paper , (33). Martins, P. S. (2020). Jobs Cronyism in Public-Sector Firms. GLO Discussion Paper , (624). Maydeu-Olivares, A., Shi, D., & Rosseel, Y. (2018). Assessing fit in structural equation models: A MonteCarlo evaluation of RMSEA versus SRMR confidence intervals and tests of close fit. Structural Equation Modeling: A Multidisciplinary Journal , 25 (3), 389-402. Moon, T., & Schoenherr, D. (2022). The rise of a network: Spillover of political patronage and cronyism to the private sector. Journal of Financial Economics , 145 (3), 970-1005. Mueller, J. E. (1970). Presidential Popularity from Truman to Johnson. American Political Science Review , 64 (1), 18-34. Nannestad, P., & Paldam, M. (1997). The grievance asymmetry revisited: A micro study of economic voting in Denmark, 1986–1992. European Journal of Political Economy , 13 (1), 81-99. Nannestad, P., & Paldam, M. (2003). The cost of ruling: A foundation stone for two theories. In E. Dorussen & M. Taylor (Eds.), Economic voting (pp. 17-44). Routledge. Neto, O. A., & Lobo, M. C. (2009). Portugal's semi‐presidentialism (re) considered: An assessment of the president's role in the policy process, 1976–2006. European Journal of Political Research , 48 (2), 234-255. 99 Table A4: Matrix of correlations. Own calculations. Board (N) Connec tion Connec tion Right Connec tion Left Connec tion (N) Total Assets Market Cap ROE Leverag e Book to Market Market Share Product ivity Returns Privatiz ed State Owners hip Right Lisbon Troika Regulat ed Board (N) 1.000 Connection 0.587 1.000 Connection Right 0.420 0.625 1.000 Connection Left 0.109 0.362 -0.167 1.000 Connection (N) 0.691 0.703 0.494 0.247 1.000 Total Assets 0.682 0.369 0.384 0.009 0.508 1.000 Market Cap 0.629 0.416 0.232 0.239 0.581 0.492 1.000 ROE 0.122 0.057 0.032 0.073 0.057 0.044 0.178 1.000 Leverage -0.018 -0.064 -0.044 -0.001 0.038 0.144 -0.142 -0.288 1.000 Book to Market -0.128 -0.102 -0.056 -0.035 -0.095 -0.041 -0.125 -0.100 0.288 1.000 Market Share 0.413 0.282 0.214 0.128 0.337 0.324 0.560 0.192 -0.217 -0.100 1.000 Productivity 0.085 0.085 -0.001 0.076 0.090 -0.215 0.224 0.100 -0.214 -0.155 0.125 1.000 Returns -0.011 -0.017 0.013 0.010 0.008 -0.008 0.095 0.196 -0.228 -0.075 0.052 0.025 1.000 Privatized 0.509 0.377 0.268 0.125 0.365 0.217 0.448 0.083 -0.060 -0.049 0.319 0.290 -0.050 1.000 State Ownership 0.253 0.240 0.079 0.244 0.290 0.061 0.361 0.044 -0.011 0.024 0.238 0.379 -0.014 0.627 1.000 Right -0.018 0.040 0.003 0.005 0.014 0.011 -0.063 -0.132 0.017 0.010 -0.042 0.000 0.107 -0.002 0.005 1.000 Lisbon 0.048 0.194 0.073 0.164 0.118 -0.099 0.092 -0.069 0.021 0.030 0.201 0.135 -0.110 0.364 0.207 0.022 1.000 Troika 0.035 -0.095 -0.032 -0.036 -0.064 -0.005 -0.010 0.007 0.121 0.128 0.064 0.000 0.058 -0.010 -0.133 0.125 -0.039 1.000 Regulated 0.549 0.413 0.352 0.128 0.427 0.496 0.451 0.044 -0.049 -0.134 0.179 0.008 -0.029 0.332 0.253 -0.016 0.093 0.016 1.000