The Erosion of the Quality of Democracy and Journalist Harassment in Latin America

During the last part of the twentieth century (between the late 1970s and the 1990s), Latin America experienced a dramatic series of democratic transitions (Hagopian and Mainwaring Reference Hagopian and Mainwaring2005). Nowadays, universal suffrage multiparty elections are virtually ubiquitous and their role in the selection of chief executives is virtually unquestioned (Zovatto Reference Zovatto2020). This is a remarkable achievement if we consider the relatively frequent episodes of democratic breakdown that characterized the region during the twentieth century (Linz and Stepan Reference Linz and Stepan1978; Mainwaring and Pérez-Liñan Reference Mainwaring and Pérez-Liñán2013), as well as the persistence of conditions that have historically challenged the survival of democracy, such as relatively high levels of socio-economic inequality (Levitsky Reference Levitsky2018). However, with few exceptions (e.g. Costa Rica, Uruguay), during the past two decades, the quality of democracy in Latin America has remained relatively low, and it has progressively deteriorated in several countries (Mainwaring and Pérez-Liñan Reference Mainwaring and Pérez-Liñán2023; Zovatto Reference Zovatto2020).

Figure 1 tracks the democratic trends experienced in Latin America during the past three decades. This graph reports regional average levels of democracy, which inevitably conceal a great deal of variation regarding the specific political regime trajectories that Latin American countries have followed over the past decades. However, the graph shows that, after having constantly progressed throughout the 1990s and early 2000s—in the wake of the so-called “third wave of democratization” (Huntington Reference Huntington1991)—the average levels of both electoral and liberal democracy in the region peaked around 2005 and then started to shrink. Remarkably, as of 2019, regional democracy levels have returned to the levels of the early 1990s.

Figure 1. Levels of Democracy in Latin America 1990–2019.

Note: The solid and dashed lines measure Latin America’s average level of democracy, based on the Varieties of Democracy (V-Dem)’s Electoral Democracy index and Liberal Democracy index, respectively (Coppedge et al. Reference Coppedge, Gerring, Lindberg and Teorell2021a). Both indexes range from 0 (min) to 1 (max).

While informative, democracy indexes summarize the democratic performance of countries or regions (as in our graph) aggregating dimensions that may perform differently. Democracy is a multidimensional form of political regime requiring formal institutions such as universal suffrage and multiparty elections, but also political rights and liberties such as freedom of association, freedom of expression, and access to alternative sources of information (Dahl Reference Dahl1971), as well as, according to some scholars (O’Donnell Reference O’Donnell2004), the establishment of the rule of law to safeguard the above-mentioned institutions, rights, and liberties.

In this regard, recent studies (Kellam and Stein Reference Kellam and Stein2016; Ponce Reference Ponce2015) have identified press freedom as a weak spot in many Latin American countries and its deterioration as a troubling trend in the region. Several sources confirm that Latin America is becoming a particularly daunting region for journalists, in particular. The crimes against journalists perpetrated by criminal organizations, especially in countries such as Colombia and Mexico, are well documented (FLIP 2019; Article 19 2022). Non-state violence and the inability of the state to protect journalists have long been a significant problem for press freedom in Latin America (Waisbord Reference Waisbord2002). However, this is only part of the story, as Latin American governments have increasingly contributed to the harassment of journalists. In 2022, in a joint statement, seven civil society organizations promoting press freedom underlined executives’ disinformation campaigns, verbal attacks, and even illegal surveillance against journalists as warnings against an ongoing repressive turn (Reporteros sin Fronteras 2022). For example, journalists have been a “favorite” target of governments in Colombia, Nicaragua, and Guatemala. Similarly, in Brazil, under Bolsonaro, harassment against critical journalists has characterized the relationship between the government and the press (Freedom House 2022). In Mexico, the recurrent attacks of López Obrador on the press recently pushed numerous journalists to sign a joint statement to denounce and rebuke the president’s attitude to the press (El Financiero 2022). In a similar vein, Human Rights Watch (2022) recently criticized López Obrador’s harassment of journalists, particularly his recurrent intimidation of critical journalists during press conferences (Human Rights Watch 2022).

As a confirmation of these concerns, Figure 2 tracks changes in the level of harassment that journalists suffer in Latin America using the Harassment of journalists indicator from the Varieties of Democracy (V-Dem) dataset. This indicator measures on a 0–4 scale the extent to which journalists are “harassed—i.e., threatened with libel, arrested, imprisoned, beaten, or killed” during their journalistic activities (Coppedge et al. Reference Coppedge, Gerring, Lindberg and Teorell2021b, 201). More specifically, a score equal to 0 refers to a situation in which, due to harassment, “no journalists dare to engage in journalistic activities” (Coppedge et al. Reference Coppedge, Gerring, Lindberg and Teorell2021b, 201), whereas a score equal to 4 indicates that, during their journalistic activities, harassment never occurs. Given that higher scores correspond to better conditions for journalists (that is, less harassment), for simplicity we relabel this indicator as Journalists’ freedom from harassment.

Figure 2. Trends in Journalist Harassment in Latin America.

Note: The solid line measures Latin America’s average level of journalists’ freedom from harassment based on the Varieties of Democracy (V-Dem)’s “v2meharjrn” indicator (Coppedge et al. Reference Coppedge, Gerring, Lindberg and Teorell2021a). We use the “linearized original scale posterior prediction” (osp) version of this indicator, which ranges from 0 (lowest score) to 4 (higher score).

As we can see in Figure 2, harassment of journalists in Latin America has worsened since the mid-2000s, as part of the more general trend of democratic erosion already depicted in Figure 1. It is worth highlighting that the observed decline is particularly severe, as it led to levels of harassment in 2019 that are worse than in 1990.

While in itself a worrisome phenomenon, harassment of journalists entails the risk of crucially distorting the functioning of democracy. Journalists are key actors in keeping the government accountable in a democratic system. According to Schedler (Reference Schedler, Schedler, Jay Diamond and Plattner1999), accountability is a multidimensional mechanism with three functions, namely, information, justification, and punishment. Justification refers to the obligation by the rulers to explain their decisions and actions. In turn, punishment concerns the power that electors and other bodies have to punish or reward the incumbent as a consequence of their actions. In this context, information should be considered as a necessary condition for this mechanism to function, as reporting news about government’s acts and decisions is essential to attribute responsibility. Hence, journalists are key accountability agents as they concur to inform citizens about what rulers do (and do not), and thus to maintaining, increasing, or decreasing a government’s popular support.

In short, while there is a debate in the literature on accountability as to how these actors should be classified,Footnote ¹ a pluralist, independent, and free press unquestionably represents an essential democratic dimension (Dahl Reference Dahl1971) and a bulwark against authoritarian-leaning rulers (Lindstaedt Reference Lindstaedt2021; Laebens and Lührmann Reference Laebens and Lührmann2021; McDevitt Reference McDevitt2022). Accordingly, the harassment of journalists inevitably contributes to the erosion of the quality of democracy. Shedding light on its causes, especially when elected governments promote it, can thus contribute to a better understanding of the declining trajectory that democracy is experiencing in Latin America and therefore represents both a theoretical and practical priority.

Declines in Governments’ Popular Support as a Determinant of Journalist Harassment

Government harassment of journalists has emerged as a recurrent and diffuse feature of Latin American politics (Reporteros sin Fronteras 2022), contributing to the decline in the average level of democracy that the region has experienced since the mid-2000s. Why do governments harass journalists though?

While the literature has rarely focused specifically on the determinants of governments’ propensity to harass journalists, it has highlighted several factors behind the restrictions rulers may impose on press freedom, more generally. Many of them can be classified in two broad categories, based on Kellam and Stein (Reference Kellam and Stein2016)’s distinction between motivating factors and enabling or obstructing factors. Motivating factors include populism (Kenny Reference Kenny2020), mass protests (Kellam and Stein Reference Kellam and Stein2016; Solis and Sagarzazu Reference Solis and Sagarzazu2020; VonDoepp and Young Reference VonDoepp and Young2013), the beginning of civil conflicts, governmental attempts to increase power through constitutional reform (VonDoepp and Young Reference VonDoepp and Young2013), economic crises, and the ideological distance between the press and a president facing a weak opposition (Kellam and Stein Reference Kellam and Stein2016).

Populist leaders have more incentives to control the press, given the attention they pay to establishing a direct, unmediated communication with voters (Kenny Reference Kenny2020). However, restrictions of press freedom can become a strategic priority also for non-populist governments during power-threatening events such as mass protests, due to the heightened impact that media coverage assumes under these circumstances (Solis and Sagarzazu Reference Solis and Sagarzazu2020; VonDoepp and Young Reference VonDoepp and Young2013). The increased relevance of controlling the press during events that are of particular importance to the government also explains the relation between press freedom and circumstances such as the beginning of civil conflicts and governmental attempts to increase power through constitutional reforms (VonDoepp and Young Reference VonDoepp and Young2013). Finally, incumbents’ decision to undermine press freedom can be motivated by the need to identify an enemy when the political opposition is weak, particularly in the presence of ideological differences between the press and the government (Kellam and Stein Reference Kellam and Stein2016).

Enabling and obstructing factors, in turn, encompass the strength of the rule of law, including judicial independence (VonDoepp and Young Reference VonDoepp and Young2016), a natural resource-based economy (Mazzuca Reference Mazzuca2013), the extent of decree powers and of institutional checks to the government, as well as the strength of the executive in congress (Kellam and Stein Reference Kellam and Stein2016). As noted by VonDoepp and Young (Reference VonDoepp and Young2016), by increasing the costs of repression, an independent judiciary can prove to be an important bulwark against the limitation of press freedom. Similarly, while stronger decree powers represent an effective instrument to dismantle press freedom, a more independent congress can provide, like judicial independence, a check to incumbents engaging in the erosion of press freedom (Kellam and Stein Reference Kellam and Stein2016).

Arguably, many of the above-reviewed factors could not only impact press freedom in general, but also, more specifically, shape rulers’ propensity to harass journalists as well as their actual ability to do so. In our view, however, popular support is another key but rarely investigated factor that could influence the relationship between rulers and journalists. For clarity, scholars have already examined how media freedom can shape citizens’ opinion about the government, either by influencing the impact of repression (Pop-Eleches and Way Reference Pop-Eleches and Way2023) or by favoring dissent (Gleditsch et al. Reference Gleditsch, Macías-Medellín and Rivera2023), as well as non-democratic governments’ attempts to manipulate the press to increase their popular support (Guriev and Tresiman Reference Guriev and Treisman2019). However, whether and how government popular support influences press freedom remains unclear. In this regard, we argue that declines in popular support could represent a motivating factor pushing executives to harass journalists, given their ability to influence public opinion. By harassing journalists through physical attacks, detentions, and threats, governments can punish and silence individuals who are perceived as contributing to their decline in public support or viewed as obstacles to regaining popularity.

Variations in the level of popular support of an incumbent leader throughout his/her mandate are routine and, in a democratic regime, they are just “part of the game”: they shape political competition and, relatedly, they influence incumbents’ chances of keeping their jobs, or else the likelihood of opposition victories and government turnover. However, the normal functioning of democratic politics should not be taken for granted. Even in regimes commonly considered democratic, executives may respond to unfavorable scenarios with practices that violate the spirit and principles of democracy.

Because decreases in popular support affect the political prospects of an executive—for instance, its ability to remain in power or implement a given policy agenda—we expect political leaders to consider and in some cases actually resort also to undemocratic strategies when their popular support declines, including the harassment of journalists. By voicing their criticism and exposing governmental wrongdoing (Brunetti and Weder Reference Brunetti and Weder2003; Solis and Antenangeli Reference Solis and Antenangeli2017), journalists can affect a government’s reputation with the public and thus with voters. At the same time, governments tend to view the control of the press and of information as an important tool to preserve their popularity (Guriev and Treisman Reference Guriev and Treisman2019). In a similar way, the repression and intimidation of journalists can be viewed by governments as a means to punish and silence individuals who are seen as contributing to their decline in public support or as obstacles to regaining popularity. Accordingly, we expect governments facing declining popular support to be more likely to target and harass journalists.

In line with our argument, Latin America offers several examples in which journalists become sensible targets when presidents lose popular support. For example, in Colombia, in the context of President Duque’s declining popular support (Valora Analitik Reference Analitik2021) and widespread popular discontent against a government-sponsored tax reform, journalists were one of the main targets of the repression orchestrated by the state during the protests of the so-called 2021 paro nacional (national strike). Following a visit to the country, the Inter-American Commission on Human Rights (CIDH 2021) highlighted severe and numerous human rights violations during the protests. The CIDH (2021, 19), in particular, underlined the attacks against journalists, who were the targets of both intimidation and physical violence. According to NGOs such as Reporters without Borders and the Foundation for Press Freedom, in such a context, security forces deliberately and systematically attacked journalists with the goal of silencing the press (Reporteros sin Fronteras 2021).

In Guatemala, based on an anti-corruption message, former comedian Jimmy Morales won around 67% of the vote in the second round of the 2015 presidential elections. However, Morales’s popular support was soon affected by the investigations of the International Commission against Impunity in Guatemala (CICIG) over his family and alleged illegal campaign financing in his presidential bid, with approval ratings plummeting below 20% in 2018 (BBC 2018). According to the Association of Guatemala’s Journalists (APG 2019, 5–6), the investigations against his family marked a turning point in Morales’s approach to the press. The APG (2019) denounced that journalists were victims of intimidation, systematic defamation campaigns, and the verbal attacks by the president, in particular. The president’s recurrent attacks on the press were condemned by the country’s ombudsman through a non-binding resolution in 2019 (Bolaños Reference Bolaños2019).

In Nicaragua, in what had been widely recognized as a non-democratic hybrid regime for years, until the wave of protests and popular dissent shaking the country in 2018, Ortega had not behaved in a particularly repressive manner. Shifter (Reference Shifter2016), for example, described Ortega’s authoritarianism as “soft” and his rule was supported by high approval ratings and a mutually convenient relationship with the business sector. However, when popular dissent broke out in 2018, Ortega engaged in harsh repression against protesters and opponents. Like Colombia, journalists were a prime target of governmental repression. As noted by officials of the Committee to Protect Journalists, the government’s response was “a systematic crackdown on freedom of expression” (Southwick and Martínez de la Serna Reference Southwick and Martínez de la Serna2021, n.p.). According to Human Rights Watch (2019), the government’s repression of journalists took a variety of forms, including the temporary closure of independent media, threats, and physical aggression.

According to the above anecdotal evidence, therefore, Latin American presidents do seem to turn against journalists as popular support decreases, as suggested by our theoretical reasoning. To test our argument in a more cogent way, we formulate the following hypothesis, which we examine in the next section through regression analysis:

The role of the popular support enjoyed by incumbents has already been highlighted as a factor contributing to the erosion of democracy, but, to our knowledge, with a main focus on how high levels of popular support can ease incumbents’ challenges to democracy (e.g., Bermeo Reference Bermeo2016; Kneuer Reference Kneuer2021; Levitsky and Loxton Reference Levitsky and Loxton2013; Mainwaring and Pérez-Liñan Reference Mainwaring and Pérez-Liñán2015; Mazzuca Reference Mazzuca2013; Velasco Guachalla et al. Reference Velasco Guachalla, Hummel, Handlin and Erica Smith2021). For instance, incumbents may take advantage of popular support to sideline other institutions or to pass institutional reforms aimed at increasing executive power and weakening checks and balances. This practice has been especially evident in Latin America, as shown, for example, by the governments of Hugo Chávez in Venezuela, Evo Morales in Bolivia, Rafael Correa in Ecuador, as well as by the recent rise to power of Nayib Bukele in El Salvador, and by several less recent cases of “delegative democracy” (O’Donnell Reference O’Donnell1994).

With few exceptions (Corrales Reference Corrales2015; Kneuer Reference Kneuer2021), however, little attention has been paid thus far to what happens when popular support drops and how such a decrease can influence an incumbent’s behavior, his/her relationship with other institutional and non-institutional actors and, in a more or less direct way, the state of democracy in a country. By examining the relationship between declines in the support enjoyed by an incumbent and his/her propensity to harass key agents of democratic accountability, such as journalists, our investigation thus complements existing research on the role of popular support in contemporary processes of democratic erosion.

Losing Popular Support and Harassing Journalists in Latin America: A Time-series Cross-sectional Analysis

When and why do Latin American governments harass journalists? In the previous section, we discussed several factors previously identified by the literature that may incentivize governments’ initiatives against journalists (and the press, more generally). We then identified popular support (and loss thereof) as a key but relatively understudied driver of presidents’ behavior against journalists, given the ability of the latter to influence citizens’ perceptions about the government and the fact that harassing journalists could be perceived by incumbents as a means to punish and silence individuals who are seen as contributing to their drop in popular support or as obstacles to regaining popularity. More specifically, we argued that declines in popular support could motivate governments’ harassment of journalists. In this section, we test this hypothesis empirically.

To this end, we conduct a time-series cross-sectional regression analysis, based on a dataset that collects information on eighteen Latin American countries (that is, all Latin American countries excluding Cuba and Haiti, due to data availability) observed from 1990 to 2019. The units of observation are country-years. We use the Journalist freedom from harassment scale which we already used in Figure 2 as a dependent variable (“v2meharjrn” in the V-Dem dataset).Footnote ² As a reminder, the higher the score, the less harassment journalists suffer. Accordingly, increases in journalist harassment (i.e., the phenomenon under investigation) correspond to declines in freedom of journalists from harassment (i.e., the indicator we use).

The main independent variable is “popular support loss,” which we measure using data from the Executive Approval Database (EAD) (Carlin et al. Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2019). EAD collects annual time-series measures of public support for political executives based on data from multiple sources (e.g., polling firms and individual public opinion polls). Specifically, we use the Net executive approval index, which ranges from −100 to +100 and measures the percentage of positive responses minus the percentage of negative responses. From this index, we derived a new Popular support loss index measuring the difference between the level of support enjoyed by an incumbent executive in year t minus the level of support the same executive enjoyed in year t-1 .

In other words, our Popular support loss index records how much support an incumbent executive loses from one year to another. Since the analysis aims to test the effect, if any, of a loss in popular support, the index is scaled so that positive values refer to support losses; a value equal to 0 indicates the absence of changes in terms of support from a year to another, and negative values are support gains. Hence, increases in this index correspond to larger losses in terms of popular support for an incumbent government. Importantly, the index records missing values for every country-year in which a new chief executive takes power, so as to avoid estimating the effect of changes in the support that two different governments enjoy (see also the table included at the end of the article as an appendix reporting descriptive statistics for all the data we use).

To test our hypothesis concerning the relationship between losses in popular support and journalist harassment, we also consider several control variables representing factors that may influence the relationship under examination in various ways. Most of the control variables that we examine are selected from the review of the literature presented in the previous section and can be thought of as factors that could either motivate, facilitate, or obstruct governments in their measures against journalists, building on Kellam and Stein’s (Reference Kellam and Stein2016) distinction.

Concerning motivating factors, we control for economic growth, proximity to elections, protests, populism, and ideology. We control for the state of the economy (using World Bank data) because executives appear more likely to turn against the press in the context of economic crises (Kellam and Stein Reference Kellam and Stein2016). For similar reasons, we use a dichotomous indicator measuring the proximity to the next elections, considering both presidential and legislative votes. This indicator takes value 1 in the year of the elections, 0 otherwise. We expect election proximity to increase the executive’s intolerance of criticism, which may in turn incentivize the propensity to attack journalists and media. Relatedly, since the literature identifies a relationship between major protests and restrictions of media freedom (Kellam and Stein Reference Kellam and Stein2016; Solis and Sagarzazu Reference Solis and Sagarzazu2020; VonDoepp and Young Reference VonDoepp and Young2013), we control for the number and intensity of mass protests per year using the Mass Mobilization Data Project (Clark and Regan Reference Clark and Regan2016)Footnote ³ and counting protests with more than 1,000 participants.

Moreover, we use a V-Dem indicator measuring the extent to which the government promotes socialism as an ideology or societal model (the higher the score, the more ideology counts). In general, we expect more ideologized political leaders to be less tolerant of pluralism among journalists and the media. Moreover, following Kellam and Stein (Reference Kellam and Stein2016), given the traditional ideological distance in Latin America between a generally conservative press and leftist governments, the left-right position of the government could be thought of as a proxy for media criticism, which in turn could motivate harassment of journalists. As another factor that could motivate harassment of journalists, we control for populism using data from the Varieties of Party Identity and Organization dataset (Lindberg et al. Reference Lindberg2022) given populists’ “Manichean” interpretation of reality (Mudde Reference Mudde2004).

Concerning enabling and obstructing factors, we control for institutional executive constraints, as measured by the Polity V “xconst” variable (Marshall and Gurr Reference Marshall and Robert Gurr2020), and opposition legislative seats as a share of total seats, using data from the Database of Political Institutions (Cruz et al. Reference Cruz, Keefer and Scartascin2021). We expect journalists to be more vulnerable in countries whose governments face few checks and balances from other institutional actors, such as the judiciary and the legislature (VonDoepp and Young Reference VonDoepp and Young2016; Kellam and Stein Reference Kellam and Stein2016). We also control for a country’s level of wealth using World Bank data on gross domestic product per capita because wealthier societies tend to be more attentive to liberties such as freedom of expression and access to alternative sources of information.

Additional control variables include political regime type, state control over territory and the raw level of support the government enjoys. Controlling for regime type is important, given the range of democratic, hybrid, and authoritarian regimes existing in Latin America. To avoid endogeneity issues deriving from a dependent variable (i.e., journalist harassment) measuring a key aspect of freedom of expression and thus of democracy (Dahl Reference Dahl1971), we use Skaaning et al.’s (Reference Skaaning, Gerring and Bartusevičius2016) Lexical Index of Electoral Democracy, which rests on a rather procedural and minimalistic definition of democracy. We measure state control of territory using the V-Dem data (“v2svstterr” indicator) to control for the presence in a country of actors different from the government which may attack journalists, given the threats that illicit organizations pose to journalists in some Latin American countries. Lastly, we also include in the model specification the raw level of popularity enjoyed by the government to account for the fact that rulers could react in different ways to drops in citizen approval depending on whether they remain relatively popular leaders or not.

We run time-series cross-sectional regression analysis on a number of alternative model specifications. Table 1 presents a selection of the analyses that we conducted. Regression coefficients are estimated using robust standard errors clustered by country. All the independent and control variables are one-year lagged with respect to the dependent variable, unless differently specified below and in the table’s notes.

Table 1. Effects of a Loss in Popular Support on Journalist Harassment

*** p<0.01, ** p<0.05, * p<0.1.

Note: Models 1–4 are estimated using robust standard errors clustered by country (in parentheses). Model 5 is estimated using panel corrected standard errors (year fixed effects omitted due to estimation problems). In Models 1–5 , all independent variables are lagged (t-1), except Election year and Protests. Model 6 is estimated without lagging the independent variables.

Model 1 is relatively simple, as it only includes our main independent variable Popular support loss, plus a lagged dependent variable (i.e., the value of the dependent variable at t-1), country fixed effects, and year fixed effects. Country and year fixed effects are admittedly atheoretical but highly recommended tools in time-series cross-sectional regression, that is, when researchers work on datasets covering multiple countries observed throughout multiple years. Country fixed effects help control for country-specific, hardly observable and time invariant factors that could influence the relationship under examination, whereas year fixed effect helps control for possible cross-national trends in how variables evolve through time. In turn, a lagged dependent variable addresses problems of autocorrelation and the fact that the measurement of the level of journalist harassment at time t could to some extent be influenced by its measurement at time t-1. Overall, these techniques are meant to maximize the accuracy of the estimated regression coefficients and thus the reliability of our findings and conclusions.

Models 2-to-4 add to the model specification the control variables described above through an incremental approach. Specifically, Model 2 includes in the regression model variables regarding the above-discussed motivating factors, namely, economic growth, populism, socialist ideology, protests (non-lagged), election year (non-lagged). Model 3 adds the previously discussed enabling and obstructing factors, namely, executive constraints, opposition seats, economic wealth. Model 4 is the full model that also includes the remaining control variables, namely, democracy, raw level of popular support, state control over territory. The last two models reported in Table 1 replicate the full model (Model 4) in alternative ways. Model 5 is run by estimating panel corrected standard errors. In Model 6, in turn, the independent variables are included in non-lagged format.

According to the analyses reported in Table 1, the empirical evidence confirms the hypothesis that losses in popular support tend to induce Latin American presidents to harass journalists and thus to have a negative effect on our Journalist freedom from harassment dependent variable. The estimated coefficients for the main independent variable Popular support loss are negative and statistically significant in all the models presented in Table 1, even when other motivating, enabling, and obstructing factors are considered, as well as after controlling for regime type, the raw level of support the government enjoys, and stateness. Concerning control variables, in turn, while the estimated coefficients are rarely statistically significant, the signs tend to be consistent with our expectations, with a few exceptions. Proximity to elections, protests, and socialist ideology negatively influence journalists’ freedom from harassment, whereas democracy, executive constraints, economic wealth, and economic growth are associated with higher levels of freedom from harassment. Counterintuitively, however, populism seems to be associated with higher rather than lower levels of journalists’ freedom from harassment, whereas stronger legislative opposition seems to correlate in a negative way.

Table 2 presents a series of robustness tests aimed to assess whether the findings from Table 1 hold to alternative measurements of our main independent variable. As reference model for this replication analysis, we use Table 1’s Model 4 (i.e., the full model). To ease comparison, the model is reported also in Table 2 as Model 7. In Model 8, we use an indicator of Popular support loss that only records declines in support above a given magnitude (0 otherwise). We establish a country-specific “above the average” threshold, in particular. As an example, if Country A, throughout the period under examination (1990–2019) displays an average level of Popular support loss equal to 5, this variable only records Country A’s values higher than 5, whereas values lower than 5 are recorded as 0.

Table 2. Robustness Checks

*** p<0.01, ** p<0.05, * p<0.1.

Note: Robust standard errors clustered by country in parentheses. The six models are run using alternative measurements of Popular Support Loss. All independent variables are lagged (t-1), except Election year and protest. Model 7 = net approval (equal to Model 4 in Table 1); Model 8 = net approval, above the average; Model 9 = relative approval; Model 10 = relative approval, above the average; Model 11 = net approval, dichotomous; Model 12 = relative approval, dichotomous.

As a further alternative, we measure Popular support loss in terms of relative approval (instead of net approval) and we present the result of this analysis in Model 9. The EAD’s Relative executive approval index ranges from 0 to 100 and measures “the percentage of positive responses divided by the sum of the percentage of positive responses and the percentage of negative responses” (Carlin et al. Reference Carlin, Hartlyn, Hellwig, Love, Martínez-Gallardo and Singer2019). Similarly to what we did in the previous models, moreover, Model 10 replicates the analysis using a country-specific “above the average” transformation of the Popular support loss variable we used in Model 9 (i.e., the one based on EAD’s Relative executive approval index).

The last two models in Table 2 are run using dichotomous indicators of Popular support loss. The dichotomous indicator we use in Model 11 is based on EAD’s Net executive approval index and takes value 1 only if the net approval for an incumbent changes from positive to negative, that is, from a situation in which a majority supports the executive (positive responses minus negative responses > 0) to a situation in which a majority does not support the executive (positive responses minus negative responses < 0). Following a similar logic, in Model 12, we use a dichotomous indicator derived from EAD’s Relative executive approval.

As we can see in Table 2, all the above-described replication analyses confirm our findings: in Latin America, declines in popular support for the government—even when measured in different ways—lead to declines in journalists’ freedom from harassment (and thus to higher levels of harassment). Overall, therefore, the analyses reported in Table 1 and Table 2 demonstrate quite unambiguously that the statistical significance of our findings is sound. However, a comparison of the models reported in Table 2—estimated using alternative versions of the main independent variable—enables us to also draw some considerations regarding the magnitude of the estimated effects. To ease the comparison, Figure 3 depicts graphically the effects on journalist harassment of a 1-unit change in the various Popular support loss measures we employed in Table 2.

Figure 3. Estimated Effect of Popular Support Loss on Journalist Harassment in Latin America.

Note: The solid lines depict the estimated coefficients of the variable Popular support loss reported in Table 2, representing the effect on journalist harassment of a 1-unit change in this variable.

In this regard, we should point out that declines in popular support have statistically significant but relatively small effects on journalist harassment. To some extent, these small-sized effects have “technical” explanations. On the one hand, for most of our variables (including the dependent variable), only annual data are available: we suspect that by focusing on smaller time periods we could obtain more precise and possibly larger estimated effects. On the other hand, we notice that the size of the estimated effects is contingent on how sensitive the scale through which we measure support for the government is to even very small changes. For instance, the −0.002 and the −0.003 effects of Popular support loss estimated in Model 7 (net approval) and Model 9 (relative approval) (Table 2 and Figure 3) correspond to 0.05% and 0.07% decreases in the degree to which journalists operate free from harassment in Latin America, respectively. However, as we measure Popular support loss by recording more substantial drops, such as country-specific “above the average” drops, the estimated effects rise to 0.10% (Model 8, net approval) and 0.18% (Model 10, relative approval) declines in the dependent variable, and the effects grow much bigger when we rely on dichotomous measures of sharper declines in popular support for the executive. The latter are estimated to have an impact on our journalists’ freedom from harassment indicator ranging from −3.07% (Model 11, net approval) to −2.51% (Model 12, relative approval).