Introduction Since his election as President of the United States, Donald Trump has repeatedly attacked the press, calling journalists “among the most dishonest human beings on earth” (Grynbaum 2017), claiming that he is in “a running war with the media” (Hirschfield Davis and Rosenberg 2017), and denouncing the press as “the enemy of the people” (Remnick 2018). He has stated that he intends to silence critics by opening up libel laws (Gold 2016), while his erstwhile chief strategist, Steve Bannon, went on record saying that the news media “should keep its mouth shut” (Grynbaum 2017). Trump is hardly alone. In fact, his actions with respect to press freedom thus far have been mild in comparison with populists elsewhere in the world. For example, In Venezuela, in January 2010, President Hugo Chávez notably clamped down on media outlets aligned with the opposition, shutting down six television news stations, including Radio Caracas Televisión (Minder 2010). In January 2018, Rappler, the Philippines news website and one of the harshest critics of President Rodrigo Duterte, had its certificate of incorporation revoked by the Securities and Exchange Commission on legally dubious grounds and has since been charged with tax evasion (BBC 2018). Perhaps most egregiously, in Turkey, in the two weeks following the failed July 15, 2016 coup, President Recep Tayyip Erdoğan shut down more than 150 news outlets, including 45 newspapers, 29 publishing houses, 23 radio stations, 16 TV stations, 15 magazines and three news agencies (Mortimer 2016); tens or even hundreds of journalists remain in prison (The New York Times 2018). A growing body of cross-national comparative evidence shows that populist rule is associated with declines in institutional checks and balances on executive authority (Houle and Kenny 2018, Huber and Schimpf 2016, Kenny 2017: ch. 2, Ruth 2018). However, in spite of the media’s theorized importance in sustaining a healthy democracy, and in spite of the threat that populists appear to pose to press freedom from cases such as those noted 1 above, we still lack a specific theory of why populist rule per se would be associated with an erosion of media freedom, whether the relationship between populist rule and press freedom might be conditioned by the ideology of the populist government in question, and a comprehensive empirical test of these relationships (Mazzoleni 2008, Mazzolini, Stewart et al. 2003). This paper argues that although the mass media has become increasingly important to all types of political parties, the way in which populist parties are organized means that they have especially strong incentives to erode press freedom and freedom of expression more broadly. In this paper, a populist party is defined by the way in which it is internally structured and linked with its supporters. This has been called the “organizational” approach to conceptualizing populism (Kenny 2017, 2019) and is closely related to the “politicalstrategic” approach (Weyland 2001, 2017). According to this approach, populism is defined as “the charismatic mobilization of a mass movement in pursuit of political power” (Kenny 2019: 1); in turn, populist parties are ones headed by charismatic leaders who seek to gain and retain power by mobilizing mass constituencies that are typically free of other political attachments. There are two reasons why populism in this sense matters for press freedom. First, while mass rallies continue to be an important means by which charismatic leaders connect with supporters, they also rely heavily on the mass media to deliver their message and mobilize voters directly. Controlling the media is thus a core objective of populists compared to other types of political leaders, who can rely on party membership, dense civil society organizations, or clientelistic linkages mediated by a network of party brokers to mobilize supporters. Second, although populists can and do have parties, they are highly personalistic; the interests of populist parties are virtually equivalent to the interests of party leaders. This 2 means that populist leaders face different time horizons and constraints on their behavior than the leaders of more deeply institutionalized parties. Unlike the latter, populists are not as motivated to ensure the survival of their party as organizations beyond their own personal political lives. Hence they are less interested in preserving the autonomy of institutions that balance governmental authority, such as the press. In contrast, leaders of non-populist parties are constrained to adhere more closely to the longer-term interests of their party. This paper thus proposes that populist parties in government are particularly likely to erode the watchdog function of the media by harassing and prosecuting critical journalists, censoring the press, shuttering opposition-aligned media outlets, and saturating the media environment through state-owned or sympathetic private media. Critically, the observed negative behavior with respect to the media is a result of populism as an organization—mobilization strategy rather than as an “illiberal” ideology per se. However, this does not imply that ideology is unimportant. On the contrary, a substantial body of research suggests that populists of the left and right should behave differently (Huber and Ruth 2017, Huber and Schimpf 2017, Katsambekis 2017, Mudde and Rovira Kaltwasser 2013, Riofrancos 2017, Rovira Kaltwasser 2012). Rather than rely on a cultural conceptualization of left and right, which might be conflated with liberal (inclusionary) and illiberal (exclusionary) policies themselves, this paper instead concentrates on the classic economic spatial dimension. It proposes that the general preference for deregulation and the autonomy of the private sector should mitigate the erosion of media freedom among right-leaning populist governments; left-leaning populist governments, in contrast, because of the greater likelihood that they will engage in the nationalization of private assets or other forms of state intervention in the market should be more likely to erode the independence of the media. 3 This paper provides a quantitative cross-national test of whether populist rule is associated with an erosion of press freedom and freedom of expression, and of whether this relationship is conditional on a party leader’s ideology. Using a dataset on populist rule in 86 countries from 1980-2014 it shows that populist government is negatively associated with the freedom of the press and freedom of expression more broadly. The substantive negative effect of populist rule on various indicators of media freedom is about a third the size of that of authoritarian rule. However, for most indicators, being on the economic right moderates the negative effect of populist rule. This paper proceeds as follows. The first and second sections discuss the conceptualizations of populism and media freedom used in this paper. The third section theorizes the relationships between party type, ideology, and the media. The fourth, fifth, sixth, and seventh sections present the data, empirical strategy, main results, and robustness checks respectively. A final section concludes. Populism as an Organizational Strategy Populism is literally a practice, system, or doctrine of the people. Scholars have variously understood this people-centric form of politics to be a type of political ideology, a set of economic policies, a manner or style of discourse, or simply a form of political mobilization (for a recent review, see Moffitt 2016). There is, of course, no true definition of populism any more than there is a true definition of democracy or justice. What matters for the purposes of theorizing the relationship between populist rule and the media and for any empirical testing of the relationship is simply that the term is clearly defined and operationalized. This paper builds on the “organizational” or “political-strategic” approach used in the early wave of political sociological studies of populism in Latin America (Di Tella 1965, 4 Germani 1978, Mouzelis 1985, van Niekerk 1974) and most recently discussed and operationalized by Kenny (2017). Populism in this sense has no core ideological content. It is simply a means by which leaders look to organize the pursuit of power in the context of a democracy. Populism is defined as “the charismatic mobilization of a mass movement in pursuit of political power” (Kenny 2019: 1). The idea that populism is a form of charismatic leadership of the masses implies that populist movements or parties have two chief characteristics that set them apart from two other pure party types, programmatic and clientelistic parties. First, authority within a populist movement or organization is arbitrary and concentrated in the person of the leader. The leader is not constrained by organizational rules and has (near) total authority over personnel and strategic decisions within the organization. Programmatic parties, in contrast, are characterized by formal rules and procedures governing the distribution of authority within the organization (Panebianco 1988), while authority within clientelistic parties is determined by factional strength, itself a result of competing groups’ ability to mobilize the most resources and blocs of clients (voters) (Schmidt, Scott et al. 1977). In addition, the lack of such arbitrary control over a movement precludes merely personalistic candidates from being coded as populist. Populism is not therefore understood here a synonym for a personalistic or independent politics, as is the case in some interpretations of Weyland (2001).1 Second, populist support is mobilized primarily through direct communication between the leader and mass constituencies in the form of traditional and social media, public rallies, and other forms of mass communication (Gurov and Zankina 2013, Kenny 2017, 2019, Weyland 2001, 2017). Following Mouzelis’s (1985) key insight, populist entities are thus distinct from parties or movements that utilize dense and persistent political 5 organizations, such as labor unions or other firmly aligned civil society organizations, on the one hand (i.e. programmatic parties), and from parties that mobilize support through a quid pro quo with supporters in which support is exchanged for particularistic material benefits networks of brokers and clients (i.e. clientelistic parties) on the other. Without both of these dimensions – authority within the party or movement and direct linkages to supporters – a leader or party is not considered populist. Populists as charismatic leaders can and do deploy rhetoric centered on “the people”; they often oppose the “elite” political “establishment” or some other minority; and they frequently undermine liberal institutional constraints on their authority (Hawkins and Rovira Kaltwasser 2018, Mudde 2004). The nature of charismatic leadership is such that we would in fact predict that such leaders should do and say some of these things to varying degrees. Appealing to a vaguely defined “people” rather than a social group with coherent economic interests makes sense as a low-cost strategy for aspiring political leaders with limited organizational resources (Kenny 2019: 19-20). A conceptualization of populism in the ideational sense could also predict a negative relationship between populist rule and press freedom. However, the advantage of the organizational approach to conceiving of populist governments in this context is that it mitigates concerns that actors are defined as populist because of their “illiberal” behavior or rhetoric, even though in practice the coding of populist governments in these two approaches turns out to be roughly similar (see the Robustness Checks Section for tests using this alternative approach). After defining press freedom in the next section, I then discuss why the two organizational dimensions of populism noted above are sufficient to make populist parties likely to erode media freedom than either programmatic or clientelistic parties, without invoking their particular beliefs or ideologies. 6 Defining Press Freedom The freedom of the press refers to the autonomy of the media from political interference or censorship. McQuail (2000, 146–7) defines media freedom as “the right to publish without any prior censorship or license and without incurring penalties, within the limits of other legal obligations.” Van Belle (2000) similarly conceptualizes a free media environment as one in which journalists can safely criticize political and economic elites at both the national and local levels. The erosion of media freedom includes a variety of actions taken by a government ranging from the harassment and prosecution of critical journalists, the censoring of press output, the closure through legal or illegal means of opposition-aligned media outlets, and the saturation of the media environment through state-owned or sympathetic private media. Contemporary techniques such as the coordinated use of internet “trolls” to drown out and threaten government critics should also count as an erosion of media freedom and freedom of speech (Tsui 2015). Freedom of expression more broadly refers to the ability of private individuals as well as those in the broadcast or print media to openly express their political views without fear of sanction. A free press, because of its ability to monitor and publicize the activities of government, especially if they are contrary to legal or ethical standards, has long been regarded as a core requirement of democracy (Kovach and Rosenstiel 2007, Mueller 1992). In the modern conception, the press is often regarded as an institutional check on the integrity of government, providing a decentralized mechanism to monitor agents (the government) on behalf of principals (the public). However, the development of a supposedly objective and politically neutral press is of relatively recent origin, consolidating in the United States only in the 1910s and ‘20s. Moreover, this situation has been eroded by the development of partisan media in recent decades (Ladd 2012). Today, some demographics reject the idea of “objective” media altogether (Marchi 2012). Yet even though the creeping advance of “fake 7 news” and outright bias undermines the role of the media in providing “objective” information to citizens (Allcott and Gentzkow 2017), a partisan media environment can still be democracy-enhancing. In this sense, what matters most is that the media environment is open and competitive. Media freedom also relates to democracy in a more diffuse sense. The freedom to express dissenting political views in public is necessary to a functioning democracy even in the minimalist sense, not to mention the fully liberal sense. If any discourse that the government finds objectionable can be censored, then opposition organization and voter mobilization become almost impossible. Populism and Press Freedom Scholars have been arguing for some time that politics has become increasingly mediatized. Mediatization refers to the increasing influence that the media has across different spheres of society (Asp 1986, Stromback 2008). In this context, mediatization means that relations between parties and voters are increasingly channeled through the mass media (Entman 1989). While in an era of high party and civil society organizational membership, supporters’ ties to their favored parties were mediated by a dense network of institutional linkages, all parties increasingly rely on direct communication through television and the print media to speak to voters. In this sense, all ruling parties have incentives to influence, if not directly control, the media. We might thus expect there to be little difference in the behavior of parties with respect to the media, at least among those in democracies. However, conceiving of populism as a type of political organization leads to distinctive predictions about the behavior of populists with respect to press freedom and freedom of expression. Even though it is true that party membership and civil society organization has declined and that programmatic parties rely increasingly on aligned media organizations to 8 maintain and increase their levels of support (Mair 2013), these parties remain distinct from populist ones in an important way. Programmatic or bureaucratic parties are deeply institutionalized and expect to outlive any particular leader or period of government. This gives such parties a markedly long time horizon, which in turn affects their propensity to maintain or undermine institutions that protect the interests of minorities (i.e., the opposition). Although any given programmatic party leader would seem to have an ostensible interest in suppressing opposed press outlets, the greater procedural constraints on leaders that exist within such parties (e.g., confidence votes) mean that the behavior of leaders adheres more closely to the interests of the party elite (if not its full membership). These party interests run against the erosion of press freedom. While programmatic parties aim to be in power continuously, because of the pluralistic organization of the party—interest group nexus, they accept the risk that they may have to survive in opposition on occasion. Any repression of the opposition while in power could lead to reciprocal repression if they themselves lose power. We would thus expect that programmatic parties should be relatively unlikely to infringe on press freedom when in power. There is relatively little written explicitly on the relationship between clientelistic parties and the media (Hallin and Mancini 2012). In theory, dependent as they are on material exchange, they should rely less heavily on the media to mobilize supporters. However, a large body of research on the media outside of Western Europe and North America suggests that the media environment itself becomes part of the spoils system (Hallin and Mancini 2012). The personalization of the state apparatus, including for example the politicized nature of the judiciary and the absence of independent watchdog institutions, allows this dominance over the media to occur in a way not possible in a programmatic system (Hallin and Papathanassopoulos 2002). Underdevelopment more broadly, itself correlated with clientelism, is in turn associated with media capture (Besley, Burgess et al. 2002). However, 9 as with programmatic parties, clientelistic parties have an institutional life-expectancy that extends beyond a single leader. The leadership of a clientelistic party also depends to a significant degree on maintaining the support of its brokers – the governors, senators, and other regional and urban bosses who control the party’s vote banks. As a result, clientelistic parties too accept the principle that they may have to survive in opposition. In some cases, in fact, rival clientelistic parties have evolved implicit agreements to rotate power and patronage between them (Blakeley 2001). While we might expect the patronage of one or other set of media outlets to change as parties in government rotate, we would not necessarily expect to see the outright repression of press freedom as this would represent a major deviation from the rules of the game. The organizational resources and electoral imperatives of populists furnish distinct expectations about how they should behave with respect to the media for two reasons. First, populists do not have the organizational depth or extensive national patronage networks to get out the vote for them. Thus, populists rely on direct communication with the masses through the media to mobilize support (Ellinas 2010). Evidence indicates that mass media access is a critical predictor of support for populist candidates (Sabatini 2012). Populists’ lack of an institutionalized support base, and dependence on the media to mobilize voters makes it more likely that they will seek to manipulate and even control the media when in office. Second, even if we assume that both populist and non-populist leaders experience similar returns from office and face a similar risk of prosecution if they lose office, populist party leaders have different time horizons from non-populist leaders, which should lead them to behave differently with respect to the media. A populist leader is interested in her own survival rather than in the continued prosperity of a party per se. Thus, once a populist has gained power, she has a greater interested in suppressing any opposition in order to retain 10 power for as long as she can. Moreover, although populists can and do have parties, they are personalistic to the extent that the interests of populist parties are equivalent to the interests of party leaders. Populist leaders are thus not constrained to act in the interests of an enduring political organization. Hence they are less interested in preserving the autonomy of institutions that balance governmental authority, such as the press. In this sense the mere structure of populist movements is sufficient to precipitate declines in press freedom and freedom of expression without invoking the putatively illiberal or anti-pluralist ideological commitments of populists per se (Müller 2016, Pappas 2015). However, this does not imply that we should neglect ideology altogether. Populist parties – like other parties – can be either left- or right- leaning. Where a populist party falls on this scale may have implications for its behavior in power (Otjes and Louwerse 2015). Indeed, it has been proposed in recent theoretical writing on populism that while right-wing populism with its “exclusionary” tendencies has anti-democratic implications, left-wing populism’s “inclusionary” posture could instead pro-democratic (Katsambekis 2017, Laclau and Mouffe 2013, Mudde and Rovira Kaltwasser 2013, Riofrancos 2017, Rovira Kaltwasser 2012). If true, then we might expect that mostly right-wing populists would erode press freedom when in power. Existing evidence in this respect is limited and ambiguous. Huber and Ruth (2017) find that participation tends to be higher under right-wing populists in government. This could be taken to indicate that right-wing populists expand the public sphere, but higher electoral turnouts could alternatively be interpreted to reflect dissatisfaction with government. While Huber and Schimpf (2017) find that right-wing populists are more likely to erode the rights of (ethnic) minorities, left- and right-wing parties are equally likely to erode general checks and balances on their authority. In this sense, we might expect there to be little 11 difference between how left- and right-wing populist parties treat establishment institutions like the media. Problematically, however, the precise understanding of left and right differs across these various theoretical and empirical works. If for example, one relies on the distinction between right-wing “exclusionary” parties and left-wing “inclusionary” parties, it could be that parties are being classified as ideologically “right-wing” precisely because they restrict freedom of expression and other rights of minorities (Mudde and Rovira Kaltwasser 2013). Any correlation between right-wing rule and restrictions on press freedom could thus be true by definition. Other research, however, more clearly dissociates the left—right dimension from a party’s stance on liberal values, instead focusing explicitly on a party’s economic policies (Huber and Ruth 2017, Huber and Schimpf 2017, Houle and Kenny 2019). I primarily follow the latter approach, testing whether there is an association either between left or right economic ideology and press freedom. While this economic conceptualization may not perfectly capture what scholars have meant when they refer to “right-wing populists” or the “radical right” in general (Mudde 2007), they are likely to be as close as we can get without defining “right” in terms of outcomes such as press freedom themselves. I theorize that right-wing populists in the economic sense should be less likely to erode press freedom than left-wing populists. Right-wing governments tend to place a higher value on the independence of the private sector in general, not simply for ideological reasons, but because of a higher dependence on donors and voters for whom market freedoms are of greater importance. Left-wing governments, in contrast, depend more on the support of low skilled workers and the poor in general; they are thus more likely to impose stiffer regulations on the private sector and to engage in the nationalization or redistribution of private assets. Evidence indicates that right-wing government orientation is associated with lower political 12 risk ratings by private ratings agencies and lower long-term bond yields, indicating strong market perceptions that right-wing governments pose lower risks for the private sector (Valentini 2015). It could, of course, be argued that the media is a special sector, but at the margins, I argue that we should be less likely to see right-wing populists erode press freedom because of the signal that any intervention in the private sector provides to the market, and hence to a right-wing government’s financial and popular support base. The following sections of this paper will test two hypotheses: H1: Populist rule is associated with declines in media freedom and freedom of expression. H2: The negative effect of populist rule on media freedom and freedom of expression is smaller the further the party/leader is to the right on economic ideology. Data Independent Variable The main independent variable is the type of party in control of the government. I rely on Kenny’s (2017) coding of populist parties from 1980-2010 and use his coding scheme to extend the coverage of the data to 2014. Geographically, the dataset covers the Americas, Asia, and Europe. Africa, much of the Middle East, and Central Asia are excluded from the dataset.2 Because we want to avoid the problem of populist rule simply being a weak proxy for authoritarianism, populist government is by definition restricted to democracy. Thus, cases in which populists have so eroded institutions that their regimes are no longer considered minimally democratic, such as that of Alberto Fujimori in Peru following the 1992 autogolpe, are not coded as populist. In this sense, any relationship between populist rule and a decline in press freedom should be considered conservative. All periods of 13 authoritarian government are coded using Cheibub, Gandhi et al. (2010) and extended through 2014. The main criteria for coding a populist party are 1.) whether the party leader is the creator of a new personalist party vehicle, or whether she substantially removed constraints on their power within the party having otherwise gained the leadership; and 2.) whether the party relied primarily on the mobilization of independent or swing voters through the charismatic appeal of the leader via personalistic appeals through the media and mass rallies rather than on institutional or clientelistic linkages in coming to power. Periods of populist government were coded on the basis of the campaign prior to a party leader taking office, which should mitigate concerns of endogeneity. Although the possibility remains that some bias exists in this retrospective coding of populist campaigners based on their subsequent behavior in office, the inclusion of leaders such as Yudhoyono in Indonesia, who is widely regarded as a moderate, indicates that this should not be the case. Based on a reading of the secondary literature and media reportage on political parties contesting elections in each country included in the original dataset, for the period from 2011 to 2014, only one additional populist government was identified, that of Otto Pérez Molina in Guatemala. Pérez Molina was a retired general who founded his own party, The Patriotic Party, in 2001. He campaigned on a “mano dura” or hardline law and order platform that often features in populist rhetoric (Kenny and Holmes 2018). In several additional cases the period of populist rule for governments in office in 2010 was extended. For example, populist rule in Bulgaria under Boyko Borisov (GERB) was extended from 2010 through to 2013, while populist rule in Hungary under Viktor Orbán (Fidesz) was extended from 2010 through to 2014. In total, this yields a set of 91 countries with a total of 195 years of full populist rule. All periods of populist rule from 1980-2014 are shown in Table A1 in the Appendix.3 14 Coalition governments in which populists are included as minority partners are not coded as populist in Kenny (2017), which tends to bias the sample of populist governments towards Latin America and Asia where populist parties have had greater outright electoral success. Presidential systems, especially those with run offs, make it more likely that a leader with a less institutionalized party can gain executive office. In Western Europe, parliamentary and proportional representation systems have made it possible for populist parties to gain substantial seat shares, but have made it harder for them to win power outright. In additional models, therefore, utilizing the coding of populist parties in Kenny (2017), I have included governments in which populist parties are junior coalition members (e.g. contributing a member of cabinet, but not the Prime Minister). These additional cases are indicated in Table A1 in the Appendix. The remaining periods of rule by non-populist democratic parties are then separated into those of clientelist and programmatic party government. Kenny (2017) codes clientelistic party links using estimates of corruption from Transparency International’s Corruption Perceptions Index as a proxy for clientelistic government. However, not only is corruption a potentially problematic proxy for clientelism, but it also shows little variation over time. I instead take advantage of richer estimates of party-voter linkages based on the v2psprlnks variable in the V-Dem dataset (Coppedge, John Gerring et al. 2017). It is based on an estimate of the main type of linkage between parties and voters, ranging from fully clientelistic to fully programmatic, with various levels of mixed clientelistic and collective or club goods based distribution in between. Any country-year observation coded as either 0 (clientelistic) or 1 (mixed clientelistic and local collective) is coded as clientelistic government; observations coded as 2-4 are coded as programmatic government. 15 To measure the economic ideology of the government in question, I utilize the Database of Political Institutions (DPI) measurement of the orientation of the executive (World Bank 2001). It is measured as a trichotomous left, right, and center variable based on the economic ideology of the leader of the party in control of the executive branch. Right parties are ones that are defined as conservative, Christian democratic, or right-wing. Left parties are ones that are defined as communist, socialist, social democratic, or left-wing. Center parties are defined as centrist or when party position can best be described as centrist (e.g. party advocates strengthening private enterprise in a social-liberal context). In the main analysis I use a continuous 3 point left-right ordinal variable. In additional robustness checks I use Right (dummy) as a dummy variable that takes on the value of 1 if a governing party is classified as right in the DPI and 0 otherwise. DPI data on economic ideology of the chief executive are missing for about a quarter of the country-year observations for which I have governing party type data. I have followed the DPI coding scheme as closely as possible in filling in gaps in the original data using the Comparative Manifestos Project, party websites, and secondary sources. The orientation of each populist chief executive is shown in Table A1 in the Appendix. Dependent variables Media freedom is a broad outcome with several constituent dimensions. The effort to limit press freedom could manifest itself in several ways, rising in intensity. These are: domination of the media environment; censorship; and harassment and intimidation. To capture these various components, I again draw on V-Dem (Coppedge, John Gerring et al. 2017).4 It includes seven variables of relevance. 16 1. Government censorship of the media (Censorship): This variable captures the extent to which the government directly or indirectly attempts to censor the print or broadcast media (V-dem variable code: v2mecenefm); 2. Government censorship of the internet (Internet): This variable estimates the degree to which the government censors internet access and content through a variety of measures. 0 is the case when there is no internet access at all, while 4 is the case when there are no restrictions (except in specified cases, such as child pornography) (v2mecenefi); 3. Range of media perspectives (Range): This variable captures whether the major print and broadcast media represent a wide range of political perspectives. 0 is only govt. perspective, and 3 is all major perspectives (v2merange); 4. Harassment of journalists (Harassment): This variable estimates the extent to which individual journalists are harassed – i.e., threatened with libel, arrested, imprisoned, beaten, or killed – by governmental or powerful nongovernmental actors while engaged in legitimate journalistic activities (v2meharjrn); 5. Media self-censorship (Self-censor): This variable estimates whether there is selfcensorship among journalists when reporting on issues sensitive to the government (v2meslfcen); 6. Media Bias against opposition (Bias): This variable captures whether there is bias in the extent and nature of media coverage in favor of the government and against the opposition (v2mebias); 7. Freedom of expression (Free expression): This variable measures the extent to which the government respects press and media freedom, and the freedom of ordinary people to discuss political matters at home and in the public sphere, as well as the freedom of academic and cultural expression (v2x_freexp). 17 For all dependent variables a positive coefficient would mean more freedom and less censorship. The pairwise correlations between the different measures of media freedom are in the range of 0.70 to 0.95 (see Table A3 in the Appendix). The disaggregation of press freedom into these various dimensions makes the V-Dem data particularly useful compared to aggregated indices (Whitten-Woodring and Van Belle 2017). Although related, each measure captures a subtly distinct aspect of press freedom; given that there is no issue of missingness in the data, the distinct measures are retained rather than collapsed into an index or latent variable so as to illuminate possibly differential effects of populist rule across different media. Control variables All models also control for time-varying socioeconomic conditions and violent conflict, which could be correlated both with political crises in which populist candidates may be successful and with the suppression of the media. To measure the state of the economy, I use the lag log of GDP per capita growth (Lag GDPPC growth), taken from the World Bank’s World Development Indicators. Conflict is measured using the UCDP/PRIO Armed Conflict Dataset’s measure of Internal Armed Conflict. Summary statistics for all variables are shown in Table A4 of the Appendix. Empirical Strategy Estimating the effect of government type on the freedom of the press presents some methodological difficulties. First, we have to control for unobserved cross national variation, which is likely to be present even when including relevant controls such as GDP per capita. To deal with this, the main models all include country fixed effects. Although the use of fixed effects results in the loss of some theoretically relevant information, the estimates can be interpreted as being on the more conservative side. I include additional random effects 18 models to examine some of the invariant institutional effects that might be associated with media freedom (see the Robustness Checks section). Second, because observations of the dependent variable (e.g., government censorship of the media) are not independent over time, we have to deal with the dynamic nature of the data; that is, the best predictor of the level of press freedom in year t0 is the level of press freedom in year t-1. This is illustrated in Figure A1 in the Appendix and shown formally in Table A5 in the Appendix, which regresses each of the measures of media freedom (described above) against their one-, two-, and three-year lags. The one-year lag is highly positively correlated with all of the dependent variables, while the two-year lag is weakly negatively correlated with it for five of the seven dependent variables. The effect of the oneyear lag is about ten times the size of the effect of the two-year lag. Although the various outcome measures are “sticky” in the sense that they move with a lag, there is no general trend in the various series. To test this more formally, I use the Fisher-type unit root tests with the augmented Dickey Fuller specification for panel data, including 1, 2, and 3 period lags with cross-sectional means removed (see Table A6). For all dependent variables with the exception of Internet censorship (itself not statistically significant in any of the main empirical models) all four of the tests reject the null hypothesis that all the panels contain unit roots at standard levels of statistical significance (p<0.05), no matter the lag structure used. Thus, except for Internet censorship, all of the dependent variable series are stationary, or I(0). To deal with the observed autocorrelation, I follow Keele and Kelly (2006) in including the one-year lag of the dependent variable in the main models. As a robustness check, I include an additional model with both one- and two-year lags (see Robustenss Checks section). I also re-run the main model with the first difference of the dependent variable (see Robustenss Checks section). All models use robust standard errors. 19 Results 20 Insert Table 1 about here Insert Figure 1 about here [title: Media freedom and freedom of expression under populist, clientelist, and authoritarian rule] Table 1 presents the main results. The coefficients for the effect of clientelist, populist, and authoritarian rule are illustrated in Figure 1. The reference category for the results is programmatic party democratic rule. The coefficient on populist rule is negative and statistically significant for most outcome measures. It is significant at the 5 percent level for Self-censorship, the Range of media views, and Freedom of expression, and at the 1 percent level for Censorship, Harassment of journalists, and media Bias. Internet censorship not significant at conventional levels. In the latter case, the coefficient is also negative, but is not statistically significant. While the latter finding is prima facie surprising, it has to be borne in mind that there are a third fewer observations available for this outcome as it is only measured from 1994 onwards. Importantly, it is worth emphasizing that populist rule is associated with a decrease in the Range of media views and an increase in media Bias. Thus, it is not the case the populists expand the range of views expressed in the public sphere. As is evident visually in Figure 1, the size of the coefficients on populist rule are about a third of the size of those for authoritarian rule. That is, while populist rule negatively affects press freedom and freedom of expression, the effect is quantitatively less than that observed under authoritarian rule. While the coefficient on Clientelist party rule is negative for each model (with the exception of Internet censorship), it is not statistically significant at conventional levels, rising only to the 10 percent level for Censor, Range, and Bias. 21 Of the control variables, Conflict is negatively associated with press freedom and freedom of expression, but the results are only statistically significant for some outcome measures: Censor, Range, Bias, and Freedom of expression. Insert Table 2 about here Next I test whether the effect of populist rule is conditional on the government’s ideological orientation. Table 2 includes an ordinal variable for the degree to which the party in control of the executive branch is on the Right. It also interacts Right with Populist to determine whether the effect of populist party rule is conditional on that party’s economic ideology. Ordinarily, in the absence of the interaction term, the coefficient for one independent variable (Populist) can be interpreted as the unique effect of that variable the outcome. A significant interaction term indicates that the effect of the predictor variable (Populism) on the response variable is different at different values of the other predictor variable (right, center, left). Including an interaction term thus allows us to examine the effect of populist party government conditional on its ideology. Populist rule is consistently associated with a decline in all measures of press freedom. Results are statistically significant at the 1 or 5 percent level for all variables, including Internet censorship. The further a party is to the Right, however, this effect is lessened for all seven dependent variables. The effect is statistically significant at conventional levels for all outcome variables, except Bias, which is significant only at the 10 percent level. These results are shown visually in Figure A2 in the Appendix. Results remain robust if Right (dummy) is coded as a dichotomous variable, taking on the value of one only if a party is of the right and zero if it is center or left (see Table A7 in the Appendix). However, the moderating effect of ideology on Internet censorship drops to significance at only the 10 percent level. I also rerun the model dropping the 633 country-year observations for which 22 DPI data on the ideology of the chief executive is missing. This reduces the number of observations to a maximum of 2,236. Table A8 shows that results are robust at standard levels of significance only for media censorship (Censor) and Self-Censorship. The effect remains in the predicted direction for Range and Harassment but is significant only at the 10 percent level. Although data limitations mean we cannot be certain why this is the case, additional evidence indicates it could be because the further a government is to the right, whether populist or not, it is more likely to be associated with protection of private sector property rights, including those of media organizations. See Appendix B for a discussion and additional tests. It could be that when we talk of “right-wing populists”, we have in mind a more cultural than economic conceptualization of what it means to be “right-wing”. I reexamined the interactive relationship between populist organization and ideology by using an alternative measure of party ideology. A government is coded as nationalist if the leader of the ruling party has a nationalist ideology according to the DPI. In Table A9 in the Appendix, I interact Populist with Nationalist. In this case, being right-wing in the nationalist sense does not have a consistent effect on media freedom or freedom of expression. While it is statistically significantly associated with a moderating effect on Self-censorship, it is associated with an augmented effect on Harassment. The interaction effect is not statistically significant for the other five dependent variables, while populism in the absence of nationalist ideology remains significant and negatively associated with all measures of media freedom except Internet censorship. The inconsistency of the interaction effect suggests that we should not over-interpret the two models that are significant. Robustness Checks 23 I first conduct additional analyses in which country fixed effects were dropped so that I could examine possible institutional covariates of populism and press freedom. Table A10 of the Appendix presents the results of a random effects model in which I include regional dummies (Western Europe and North America, Latin America and the Caribbean, and Asia) and additional non-time-varying institutional controls. GDP per capita at constant 2010 US dollars is taken from the World Bank’s World Development Indicators and is lagged and logged (Log GDPPC); to measure party system instability, which we might expect to be correlated with both populists coming to power and with weaker political oversight of executive behavior, I use the lag of the fragmentation of the party system (Lag party fractionalization). It is calculated as the probability that two deputies picked at random from the legislature will be of different parties. It is taken from the DPI. Presidentialism is taken from Robert Elgie (2018). Populist rule remains significant for all outcome measures (except Internet). In contrast to the fixed effect models, Clientelist government is statistically significantly associated with declines in freedom for five of the seven outcomes. This suggests that there may be some omitted country-level factor that is both associated with the prevalence of clientelism and with lower protections for press freedom, but which are not captured by a simply measure of development (Lag Log GDPPC is not significant). Regions have no consistent effect on press freedom or freedom of expression. Asia is negatively associated with an increase in Internet censorship and Harassment of journalists. Censorship, Harassment, and Self-Censorship are lower in Western Europe and North America. I next include several tests to address possible confounding due to the dynamic nature of the data. I retest the main models (Table 1) with the inclusion of a year dummy, a linear year trend, and a quadratic year trend (see Tables A11 – A13 in the Appendix). I also re-run the main models with both 1-year and 2-year lags (see Table A14 in the Appendix). Lastly, I 24 re-run the main model with the first difference of the dependent variable (see Table A15 in the Appendix). Results remain robust. To address the possibility of reverse causation, in which press freedom might affect the likelihood of populist rule, I use two approaches. First, I use a panel vector autoregression model and then conduct a Granger causality test to examine whether either populist rule is “Granger-caused” by any of the independent variables (Abrigo and Love 2015). The crosssectional mean of each variable is subtracted from each variable in the model before estimation to remove any extraneous time effects. To ensure that the panel is strongly balanced, panels for which there were not the full 35 years of observations were dropped (301 observations in total). As the table A16 in the Appendix indicates, there is no evidence of Granger causality. Second, I examine reverse causality in a more substantive sense by testing whether the lag of any measures of press freedom are associated with the likelihood of a populist coming to power in a multivariate regression model. There only variable with a weak association is media bias. However, it is not significant at conventional levels (p>0.05). See Table A17 in the Appendix. To mitigate the possibility that the results are being driven by choices around the coding of populist governments, I reestimate the main models with a number of variations. First, I re-run models 1-7 of Table 1 excluding each populist in turn. Results remain robust (see Tables A18 to A24 in the Appendix). Second, I re-code populist governments utilizing the ideational rather than organizational conceptualization of populism using the dataset of Ruth (2018), which covers Latin America from 1979 to 2014. Results are robust to this alternative conceptualization of populism as shown in Table A25 in the Appendix. Populist government is negatively 25 associated with all measures of press freedom and freedom of expression, and is statistically significant except in the case of the Internet censorship. Ruth’s coding has the advantage of being based on the nature of a candidate prior to coming to office. However, to the extent that a particular party’s (populist) ideology may have been revealed through rhetorical attacks on the press or even by illiberal behavior in general when in government, there is a possibility that any negative empirical relationship would be true by definition. This is arguably unlikely, but it remains the case that the “organizational” approach provides the more conservative test. I next consider whether the effect of populist rule is present only in cases where populists hold the main executive position in government or also for cases in which populists are minority coalition partners. Results remain robust to the inclusion of populist parties as coalition members, when conceptualized in the organizational sense (Table A26 in the Appendix). However, the organizational approach tends to exclude some ideologically antiestablishment, but otherwise programmatically structured, parties in Western Europe (Mudde 2007). I thus also use Huber and Ruth’s (2017) coding of populist parties as minority coalition members based on the ideational coding of populism. Results are not robust to this coding of populist coalition governments (see table A27 in the Appendix). This result is not surprising as this dataset is restricted to Western Europe, where formal institutional protections for the press and freedom of expression are likely to be higher than for less consolidated democracies in Latin America and Asia. Importantly, however, it suggests that populism in an ideological sense is not sufficient to pull a non-populist majority coalition partner in the direction of repressing media freedom, at least when other institutions (e.g., the judiciary) retain their autonomy. Conclusion 26 This paper makes several contributions to an emerging empirical research agenda on the effects of populism. First, it provides a novel causal mechanism grounded in the organizational structure of populist parties that links populist party rule to the repression of press freedom and freedom of expression. Second, it adds to a developing body of crossnational research which documents the negative implications of populist party government on the rule of law, electoral quality, and other liberal democratic institutions (Houle and Kenny 2018, Huber and Schimpf 2016, Kenny 2017). Third, by examining the interaction between populism as an organizational form and left-right party ideology, the paper has shown that even though populism per se matters for press freedom and freedom of expression, its effects can be mitigated (exacerbated) by a party’s economic orientation. Whether understood as a type of mobilization strategy, a style of politics, or an ideology, populism has often proved to be a frustratingly capacious category. By including actors with often quite divergent policy preferences, it has often been difficult to say what exactly populism itself might be doing. Some populists stress opposition to immigrants or other ethnonational minorities; others instead rail against the economic elite. We might well expect these actors to behave quite differently with respect to freedom of expression. This paper, however, finds that all populists restrict press freedom, even if being on the economic right moderates this relationship. This latter finding, moreover, suggests that future research might benefit from examining how the effects of populism are conditioned by the particular electoral coalitions and party donors on which different populists rely rather than on populists’ ideology per se. These findings have critical implications for our theoretical understanding of the relationship between populism and democracy. One of the mainstay assumptions of contemporary democratic theory is that press freedom has a particularly important role in 27 maintaining democracy. Indeed, to qualify as minimally democratic, a government must allow the opposition to freely contest elections (Przeworski 1999). If political views that the government finds objectionable can be censored, then opposition organization and voter mobilization becomes almost impossible. In this context, press freedom would seem to be essential to populists’ claims of democratic legitimacy; without freedom of the press and freedom of expression more broadly, a populist political leadership cannot claim to represent public opinion; rather it can only claim to dictate it. 28 References Abrigo, Michael R. M., and Inessa Love. 2015. “Estimation of panel vector autoregression in Stata: A package of programs.” Paper presented at the Panel Data Conference, Central European University, Budapest. http://paneldataconference2015.ceu.hu/Program/Michael-Abrigo.pdf. Allcott, Hunt and Matthew Gentzkow (2017). “Social media and fake news in the 2016 election.” Journal of Economic Perspectives 31(2): 211-236. BBC (2018). “Philippine news website Rappler has licence revoked by SEC.” BBC News. January 15. https://www.bbc.com/news/world-asia-42692723 Besley, Timothy, Robin Burgess, and Andrea Prat. 2002. “Mass Media and Political Accountability.” In The Right to Tell: The Role of Mass Media in Economic Development, edited by Roumeen Islam, Simeon Djankov and Caralee McLeish, 4560. Washington DC, The World Bank. Blakeley, Georgina. 2001. “Clientelism in the building of state and civil society in Spain.” In Clientelism, interests, and democratic representation: The European experience in historical and comparative perspective, edited by Simona Piattoni, 77-100. New York: Cambridge University Press. Cheibub, José A., Jennifer Gandhi and James R. Vreeland. 2010. “Democracy and dictatorship revisited.” Public Choice 143(1): 67-101. Coppedge, Michael, John Gerring, Staffan I. Lindberg, Svend-Erik Skaaning, Jan Teorell, David Altman, Michael Bernhard, Steven Fish, Adam Glynn, Allen Hicken, Carl Henrik Knutsen, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Pamela Paxton, Daniel Pemstein, Laura Saxer, Brigitte Seim, Rachel Sigman and J. Staton (2017). “V-Dem Codebook v7.” Varieties of Democracy (V-Dem) Project. 29 Di Tella, Torcuato. 1965. “Populism and Reform in Latin America.” In Obstacles to Change in Latin America, edited by Claudio Veliz, 47-73. London: Oxford University Press. Elgie, Robert. 2018. “List of Presidential, Parliamentary, and other Countries.” http://www.semipresidentialism.com/?p=195 Ellinas, Anotonis A. 2010. The media and the far right in western Europe: Playing the nationalist card. Cambridge: Cambridge University Press. Entman, Robert M. 1989. Democracy without citizens: Media and the decay of American politics. New York: Oxford University Press. Germani, Gino. 1978. Authoritarianism, fascism, and national populism. New Brunswick, N.J.: Transaction Books. Gold, Hadas. 2016. “Donald Trump: We're going to 'open up' libel laws.” Politico. 26 February. https://www.politico.com/blogs/on-media/2016/02/donald-trump-libellaws-219866. Grynbaum, Michael M. 2017. “Trump Strategist Stephen Bannon Says Media Should ‘Keep Its Mouth Shut’.” The New York Times. 26 January. https://www.nytimes.com/2017/01/26/business/media/stephen-bannon-trump-newsmedia.html. Gurov, Boris and Emilia Zankina. 2013. “Populism and the Construction of Political Charisma: Post-Transition Politics in Bulgaria.” Problems of Post-Communism 60(1): 3-17. Hallin, Daniel C. and Paolo Mancini. 2012. Comparing media systems beyond the Western world. New York: Cambridge University Press. Hallin, Daniel. C. and Stylianos Papathanassopoulos. 2002. “Political clientelism and the media: Southern Europe and Latin America in comparative perspective.” Media, Culture & Society 24(2): 175-195. 30 Hawkins, Kirk and Cristóbal Rovira Kaltwasser (2018). “Introduction - Concept, Theory, and Method.” In The Ideational Approach to Populism: Concept, Theory, and Analysis, edited by Kirk Hawkins, Ryan E. Carlin, Levente Littvay and Cristóbal Rovira Kaltwasser, 1-24. New York, Routledge. Hirschfield Davis, Julie and Matthew Rosenberg. 2017. “With False Claims, Trump Attacks Media on Turnout and Intelligence Rift.” The New York Times. 21 January. https://www.nytimes.com/2017/01/21/us/politics/trump-white-house-briefinginauguration-crowd-size.html. Houle, Christian and Paul D. Kenny. 2018. “The Political and Economic Consequences of Populist Rule in Latin America.” Government and Opposition 53(2): 256-287. Huber, Robert A. and Saskia P. Ruth. 2017. “Mind the gap! Populism, participation and representation in Europe.” Swiss Political Science Review 23(4): 462-484. Huber, Robert A. and Christian H. Schimpf. 2016. “Friend or Foe? Testing the Influence of Populism on Democratic Quality in Latin America.” Political Studies 64(4): 872-89. Huber, Robert A. and Christian H. Schimpf. 2017. “On the distinct effects of left-wing and right-wing populism on democratic quality.” Politics and Governance 5(4): 146-165. Katsambekis, Giorgos. 2017. “The Populist Surge in Post‐Democratic Times: Theoretical and Political Challenges.” The Political Quarterly 88(2): 202-210. Keele, Luke. and Nathan J. Kelly. 2006. “Dynamic models for dynamic theories: The ins and outs of lagged dependent variables.” Political Analysis 14(2): 186-205. Kenny, Paul D. 2017. Populism and Patronage: Why Populists Win Elections in India, Asia, and Beyond. Oxford: Oxford University Press. Kenny, Paul D. 2019. Populism in Southeast Asia. New York: Cambridge University Press. 31 Kenny, Paul D. and Ronald Holmes. 2018. “A New Penal Populism? Rodrigo Duterte, the War on Drugs, and Public Opinion in the Philippines.” Paper presented at the World Congress of the International Political Science Association, Brisbane. Kovach, Bill and Tom Rosenstiel. 2007. The elements of journalism: What newspeople should know and the public should expect. New York: Three Rivers Press. Laclau, Ernesto and Chantal Mouffe. 2013. Hegemony and socialist strategy: Towards a radical democratic politics. London: Verso. Ladd, Jonathan M. 2012. Why Americans hate the media and how it matters. Princeton: Princeton University Press. Mair, Peter. 2013. Ruling the void : the hollowing of Western democracy. London: Verso. Marchi, Regina. 2012. “With Facebook, blogs, and fake news, teens reject journalistic “objectivity”.” Journal of Communication Inquiry 36(3): 246-262. Mazzoleni, Gianpetro. 2008. “Populism and the Media.” In Twenty-first century populism: The spectre of Western European democracy, edited by Daniele Albertazzi and Duncan McDonnell, 15-29. Basingstoke: Palgrave Macmillan. Mazzolini, Gianpetro, Julianne Stewart and Bruce Horsfield (Eds.). 2003. The media and neo-populism: A contemporary comparative analysis. Westport, CT: Praeger. Minder, Benedict. 2010. “Chávez closes down opposition media outlets.” Financial Times. 25 January. https://www.ft.com/content/5b28989a-0905-11df-ba88-00144feabdc0. Moffitt, Benjamin. 2016. The global rise of populism: Performance, political style, and representation. Palo Alto: Stanford University Press. Mortimer, Caroline. 2016. “President Erdogan's crackdown on the Turkish media since the coup in one chart.” The Independent. 28 July. https://www.independent.co.uk/news/world/europe/turkey-coup-attempt-presidenterdogan-crackdown-turkish-media-chart-military-a7160381.html. 32 Mouzelis, Nicos. 1985. “On the Concept of Populism: Populist and Clientelist Modes of Incorporation in Semiperipheral Polities.” Politics & Society 14(3): 329-348. Mudde, Cas. 2004. “The populist zeitgeist.” Government and Opposition 39(4): 542-563. Mudde, Cas. 2007. Populist radical right parties in Europe. Cambridge: Cambridge University Press. Mudde, Cas and Cristóbal Rovira Kaltwasser. 2013. “Exclusionary vs. inclusionary populism: Comparing contemporary Europe and Latin America.” Government and Opposition 48(2): 147-174. Mueller, John. 1992. “Democracy and Ralph's pretty good grocery: Elections, equality, and minimal human being.” American Journal of Political Science 36(4): 983-1003. Müller, Jan-Werner. 2016. What is Populism? Philadelphia: University of Pennsylvania Press. Otjes, Simon and Tom Louwerse. 2015. “Populists in Parliament: Comparing Left-Wing and Right-Wing Populism in the Netherlands.” Political Studies 63(1): 60-79. Panebianco, Angelo. 1988. Political parties: Organization and power. Cambridge: Cambridge University Press. Pappas, Takis. 2015. “Modern Populism: Research advances, conceptual and methodlogical pitfalls, and the minimal definition.” In Oxford research encyclopedia of politics. DOI: 10.1093/acrefore/9780190228637.013.17. Przeworski, Adam. 1999. “Minimalist conception of democracy: A defense.” In Democracy's Value, edited by Ian Shapiro and Casiano Hacker-Cordón, 23-55. New York, Cambridge University Press. Remnick, David. 2018. “Trump and the Enemies of the People.” The New Yorker. 15 August. https://www.newyorker.com/news/daily-comment/trump-and-the-enemies-of-thepeople. 33 Riofrancos, Thea. 2017. “Democracy Without the People: Left populism vs. incipid pluralism” n+1. https://nplusonemag.com/issue-28/politics/democracy-without-thepeople-2/. Rovira Kaltwasser, Cristóbal. 2012. “The ambivalence of populism: Threat and corrective for democracy.” Democratization 19(2): 184-208. Ruth, Saskia P. 2018. “Populism and the Erosion of Horizontal Accountability in Latin America.” Political Studies 66(2): 356-75. Sabatini, Fabio. 2012. “Who Trusts Berlusconi? An Econometric Analysis of the Role of Television in the Political Arena.” Kyklos 65(1): 111-131. Schmidt, Steffen W., James C. Scott, Carl H. Landé and Laura Guasti (Eds.). 1977. Friends, followers, and factions: A reader in political clientelism. Berkeley: University of California Press. The New York Times. 2018. “Turkey Sentences 24 Journalists to Prison, Claiming Terrorism Ties.” 9 March. https://www.nytimes.com/2018/03/09/world/europe/turkeyjournalists-prison-terrorism.html. The World Bank. 2001. Database of Political Institutions. Tsui, Lokman. 2015. “The coming colonization of Hong Kong cyberspace: Government responses to the use of new technologies by the umbrella movement.” Chinese Journal of Communication 8(4): 1-9. Valentini, Marina. 2015. “The Impact of Party Orientation on Political Risk and Foreign Direct Investment Flows.” Senior Honors Thesis, New York University. Van Belle, Douglas A. 2000. Press Freedom and Global Politics. Westport, CT: Praeger. van Niekerk, A. E. (1974). Populism and Political Development in Latin America. Rotterdam: Rotterdam University Press. 34 Weyland, Kurt. 2001. “Clarifying a contested concept: Populism in the study of Latin American politics.” Comparative Politics 34(1): 1-22. Weyland, Kurt. 2017. “Populism: A Political-Strategic Approach.” In The Oxford Handbook of Populism, edited by Cristóbal Rovira Kaltwasser, Paul Taggart, Paulina Ochoa Espejo and P. Ostiguy, 48-72. New York: Oxford University Press. Whitten-Woodring, Jenifer and Douglas A. Van Belle. 2017. “The Correlates of Media Freedom: An Introduction of the Global Media Freedom Dataset.” Political Science Research and Methods 5(1): 179-188. 35 Acknowledgments I presented an early version of this paper at the School of Politics and International Relations at ANU; I would like to thank attendees for their comments. I would also like to thank Duncan McDonnell for his comments and suggestions on a previous draft of this manuscript. 36 Declaration of Conflicting Interests The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. 37 Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. 38 Notes 1 For instance, while Vargas Llosa, a personalist presidential candidate in Peru is coded as populist in Weyland (2001), he is not coded as such in Kenny (2017), as he lacked control over a political movement that was independent of the conservative parties that supported his candidature. 2 This geographical coverage seems to have been dictated by the availability of the independent variables used to explain the electoral success of populist candidates in Kenny (2017). Table A2 in the Appendix lists the states included. Temporally, the coverage is intended to capture populists during the long “third wave” of democratic transitions. 3 If a populist party takes office prior to July 1 in a given year, that year is coded as populist. If it comes to power after July 1, that year is coded as non-populist; populist government begins in the following year. 4 The main alternative measures are the Reporters without Borders Press Freedom Index and Freedom Houses’ Index of Press Freedom. The former runs only from 2002 so would yield a dataset only about a third of the size of the one currently used. The latter index has changed in composition over time and is thus unsuitable for panel data analysis. 39