“It’s the Economy, Stupid!”, Is it not? The Relationship between Press Freedom and the Status of the Economy in Western Media Systems

ABSTRACT

The existing literature presents several studies which show that the levels of press freedom can affect the state of the economy. What has not thus far been investigated is whether the economy can affect the levels of press freedom, the specific economic conditions that mainly affect it and the differences among various countries. This study attempts to analyze the impact of economic conditions on the degree of press freedom, in 18 countries of the western world based on a quantitative analysis for the period 2002–2019, and advance our understanding of this relationship. We find that the state of the economy within a country can affect the level of press freedom while the effects of economic conditions on the degree of press freedom seem to vary among different media systems of the western world.

KEYWORDS:

• Press freedom
• economy
• media systems
• journalists
• western countries
• quantitative analysis

In 1992 James Carville coined the phrase “It’s the economy, stupid” as one of the Clinton campaign’s key political messages. Although Carville was not referring to the relationship between press freedom and the economy, today that message appears quite timely as the status of the economy seems to affect multiple aspects not only of the socio-political environment, but also of fundamental, constitutional human rights.

Press freedom goes hand in hand with the level of democracy: the higher a society’s democratic index, the higher its level of press freedom (Sussman 2003). Political and legal factors that can impact on press freedom have been extensively assessed since the seminal work of Siebert, Peterson, and Schramm (1956) (e.g., Weaver 1977; Curran 1996; McQuail 2000/2010) and a series of influential studies in recent years (e.g., Norris 2006; Becker, Vlad, and Nusser 2007; Oster 2015; Graber 2017; Hallin 2020) have examined different conceptual approaches to categorizing threats to press freedom.

We also know that democracy and press freedom tend to increase when the level of corruption is low (Camaj 2013; Hamada, Abdel-Salam, and Elkilany 2019). Recently, we have seen that the levels of press freedom can affect the state of the economy (Nguyen et al. 2021). What has not thus far been investigated is how deeply the economy can affect levels of press freedom, the specific economic metrics that mainly impact it and how these correlations may differ among different countries of the western world. In other words, whether there is a relationship between press freedom and the economy, the reverse of the one already examined in recent years.

This study attempts to advance our understanding of this relationship and analyze the impact of economic conditions on the degree of press freedom. The research is based on a quantitative analysis of data for the period 2002–2019 for 18 countries of the western world. There are two central questions that guide the study overall: First, to what extent do economic conditions impact on press freedom in democratic countries of the western world, and, second, whether there are any differences in the way(s) the variation of economic conditions affects the degree of press freedom among western countries.

Press Freedom and the Economy: Current Trends and Characteristics

In recent years, the proportion of the global population enjoying a free press has been steadily declining especially in western countries (RSF 2022). The effects the economy may have on press freedom have been identified several years ago by Freedom House (2004), one of the main international organizations, along with Reporters Without Borders, which assess and measure press freedom around the world.

The Great Recession after 2008 seems to offer a solid ground for testing this argument, as one by one the countries of the European South started to impose strict measures to support their national economies, although the crisis presented different characteristics in the countries it affected. However, all the crisis-ridden countries seem to have experienced a gradual decline in the amount of freedom their media enjoyed. For example, between 2010 and 2015, a period with severe economic obstacles for the European South, Greece and Italy present two of the most typical characteristics: Greece ranked 70th in the 2010 press freedom index and dropped to 71st in 2012 and 91st in 2015; Italy ranked 49th in 2010 and dropped to 61st in 2012 and 73rd in 2015 (RSF 2010/2012/2015). Throughout this period, both countries suffered tax rises and strict measures to cover their economic deficit. At the same time, their media endured widespread staff cutbacks and closures of outlets, while journalists faced substantial pressure from media owners and politicians (Masrorkhah and Lehnert 2016). Against this background, this study draws from both economic and media scholarly work to assess the impact of economic conditions on press freedom in different media systems of the western world. Although these countries mostly enjoy a democratic political status, their political and media systems are quite different, and their economic status also varies significantly, as the Great Recession after 2008 disproportionally affected nation-states of the western world.

Several studies have assessed the relationship between press freedom and economic conditions, although from very different points of view. The vast majority focus on the effects of press freedom on the economy and economic development (e.g., Nguyen et al. 2021). However, the impact of economic conditions on press freedom has not thus far been the focus of systematic research, to the best of our knowledge, although several studies have individually examined various macro-economic variables in relation to press freedom.

For example, Dutta and Roy’s work (2009) established FDI (Foreign Direct Investment) as a major determinant of press freedom. Their study showed that global integration can strengthen the media sector financially, make it technologically enhanced and can also improve the economic environment as a whole. This, in turn, would work towards the enhancement of press freedom. Alam, Z, and Shah (2013) analyzed the relationship between press freedom, FDI and economic growth using a balanced panel of 115 countries. They established the existence of a bidirectional relationship between press freedom and economic growth, and FDI and press freedom in the short and long run, indicating that an economically growing country implements additional press freedom, while the flow of FDI into the economy may be an important potential channel through which press freedom and economic growth are related.

Of the many economic indicators, the major objectives of an economy are high level and rapid growth of output, low levels of unemployment and stable prices (low and stable inflation). Economic growth is a measure of expansion of the economy over time and is measured by GDP per capita (Samuelson and Nordhaus 2009). It is important because growth indicates that the economy’s ability to produce goods and services is rising (Mankiw 2018). Also, higher savings can help finance higher levels of investment and boost productivity over the longer term (OECD 2021). Unemployment has a dual impact: economic (economic losses are the greatest documented wastes in a modern economy) and social (the emotional and psychological impact of long periods of unemployment). One of the instruments used for macro-managing the economy is government expenditure, which influences the relative size of spending, private savings and consumption (Samuelson and Nordhaus 2009). Debt is a key indicator of the sustainability of government finance. Consecutive large fiscal deficits are strongly detrimental to the sustainability of public finances, as they are financed by additional debt. When the level of outstanding debt is high, the cost of servicing that debt pushes a country further into deficit, thereby hindering fiscal sustainability (OECD 2021). In recent years, one of the most important variables that has generally implied increased economic activity for a country is the ten-year-bond rate, which is deemed an important anchor point for investors as a “risk-free rate of return” that is used in valuation models to calculate the value of assets, including shares, property, infrastructure and fixed interest movements; lower bond rates result in higher asset values and, by contrast, higher bond rates result in lower asset values (Draper 2021).

This study builds on the previous works of Tran et al. (2011) and Bjørnskov (2018). Tran et al. (2011) identified that the link between overall human development (including economic growth, gross private capital flows and FDI net flows) is not established conclusively and concluded that press freedom is a better predictor of development than development is of press freedom (p.185). Bjørnskov’s (2018) work shows that improvements in economic freedom are associated with subsequent improvements in press freedom. Bjørnskov’s starting point is the hypothesis put forward by Nobel Laureates Hayek (1944) and Friedman (1962), which states that economic freedom is causally associated with stable democracy; from there, Bjørnskov moves on to test whether changes in economic freedom result in changes in press freedom.

This work seeks to test the following hypotheses:

H1. The state of the economy within a country can affect the level of press freedom.

H2. The effect of economic conditions on the degree of press freedom varies among different media systems.

To assess the degree to which economic conditions may impact on press freedom in different countries/media systems, it is important to take into consideration the level of corruption these countries are dealing with, in line with previous research on this area. Corruption—interpreted as the abuse of public power for private gain (Liu 2016)—is widely acknowledged to have negative effects on economic growth and economic conditions (Shen and Williamson 2005). Hamada, Abdel-Salam and Elkilany’s (2019) empirical investigation finds evidence of a significant relationship between higher levels of press freedom and lower levels of corruption (see also Camaj 2013). In addition, Nam (2012) argues that there are theoretical as well as empirical reasons to suspect the reverse causation between press freedom and corruption. The two-way causality was identified by Besley and Prat (2006) and Djankov et al. (2003). As such, this work includes corruption in the variables examined, to test possible effects on press freedom and examine this reverse causation in the countries examined here, in line with Hamada (2020), who included economic conditions in his hierarchical universal theory of journalism-corruption determinants.

Moreover, in addition to corruption, the existing literature refers to the legal framework for accessing information as one of the most notable variables that can affect press freedom. According to Vleugels (2009), the working operational definition of Freedom of Information Legislation includes law in a strict sense, a legal guarantee for right to access, complaint and appeal possibilities, and national level legislation. Nam (2012), building on earlier studies in the field (Banisar 2006) found that the impact of the legislation on press freedom is determined by control of corruption, the type of political regime (democratic or authoritarian), and the status of the national economy. Following this line of research, and in order to capture the broader socio-political context of this variable, we chose to use in this study broader data referring to: (a) liberal democratic rights, such as the importance of protecting individual and minority rights against the state, strong rule of law, independent judiciary, and effective checks and balances that limit the exercise of executive power; (b) civil liberties constituted, among others, by the absence of constraints of private liberties and political liberties by the government; and (c) political civil liberties understood as freedom of expression and association. Data used derive from the V-Dem database and are analytically assessed in the Method section.

Assessing Press Freedom in Different Media Systems of the Western World

In examining the relationship between press freedom and economic conditions, previous studies (e.g., Tran et al. 2011) establish the argument that the link between the two may depend on the way press freedom is defined in each society. This may well be the case when examining press freedom in relation to the level of democracy different countries tend to enjoy. For example, Gunaratne (2002), as well as several other scholars examining press freedom (e.g., Becker, Vlad, and Nusser 2007; Graber 2017), adopt the Freedom House (2010/2019) classification of nation-states around the world as “free, partly-free and not-free”. However, this study attempts to investigate the relationship between press freedom and the economy in western countries (mainly European ones and the US), which all belong to the category of free countries; as such the Freedom House classification is of limited use here. In turn, we consider the fundamental theory of Hallin and Mancini (2004) more appropriate for the needs of this study, as we explain in this section.

Hallin and Mancini (2004) introduced three types of media systems in the western world, namely: the Mediterranean/polarized pluralist (which prevails in the countries of Southern Europe), the North European Democratic/corporatist (which prevails across northern continental Europe) and the North Atlantic/Liberal model (which prevails across the UK, Ireland, Canada and the USA). They suggested that media systems can be classified based on four key dimensions: the degree of state intervention in the media, the extent of political parallelism, the development of media markets, and the level of journalistic professionalism.

Notwithstanding, since 2004 Hallin and Mancini’s classification received a great deal of criticism, mainly as regards the dimensions used to classify media systems (e.g., Norris 2009), the effect of traditional media on media systems’ characteristics (Hardy 2012), and more recent socio-economic developments that caused a series of transformations in the media landscape (e.g., Bruggemann et al. 2014). Indeed, since 2004, these media systems have experienced several transformations, which Hallin and Mancini (2017) themselves have recognized, acknowledging that media systems tend to undergo variation based on external and/or internal factors. In fact, Humprecht et al. (2022) argue that there have been significant changes in the digital age as a result of the development of information technologies, the most notable one being the shrinkage/disappearance of the liberal model. In their study, they include 30 countries and several new indicators (such as media freedom and the use of online news). What all these studies have in common is that while they present strong arguments for considering possible advancements in the initial theory of Hallin and Mancini (2004), none of them considers the theory outdated; rather they tend to build on Hallin and Mancini’s initial typology by examining different dimensions/parameters to classify media systems.

We deem it appropriate for the needs of this study to use the initial theory as a solid categorization criterion, as it allows us to dig deeper into the relationship between press freedom and economic conditions in different media systems and extend our findings in relation to economic dimensions, beyond those already examined. To the best of our knowledge, although previous studies in the field (as earlier presented in this section) identify points of dispute with the initial theory, the relationship under scrutiny here has not thus far been examined on the level of media systems. As economic conditions tend to differentiate among the different types of media systems, especially after the Great Recession of 2008, their impact on press freedom is expected to vary and may offer a new approach in assessing media systems of the western world.

In line with more recent scholarship in the field (e.g., Voltmer 2008), we added a fourth type of media system (the post-Communist countries) to extend our findings beyond Western Europe and the US, and include democratic countries of Central and Eastern Europe. We based this addition on the argument that socio-political changes that took place after 1989 in several Eastern European countries pointed towards the emergence of post-Communist media systems in what Voltmer (2008) calls “the “new’ democracies” (see also analysis in Norris 2009; Maniou 2022). Bearing these in mind, we will discuss our findings in relation to more recent scholarship in the field of western media systems and on the basis of a model of media systems categorization that flows from economic conditions to press freedom.

Method and Sample

The study is based on a quantitative analysis of data for the period 2002–2019 for 18 countries. The countries selected for the study constitute examples of each one of the four types of media systems in the western world, following Hallin and Mancini (2004), Voltmer (2008) and Hallin and Mancini (2013). As such, beyond the ones that constitute typical examples for each category as examined by Hallin and Mancini (2004 and 2013) (Netherlands, Sweden, Norway, Finland, Denmark, the USA, the UK, Ireland, Canada, Italy, Portugal, Spain, Greece), we added the Czech Republic, Bulgaria, Romania, Poland and Hungary as examples of the post-Communist countries.

The categorization of countries into the four groups of the study is presented in Table 1.

Table 1. The groups of countries (national media systems) used for the study.

Groups	Countries
Group 1: Democratic/corporatist media systems	Netherlands, Sweden, Norway, Finland, Denmark
Group 2: Liberal media systems	USA, UK, Ireland, Canada
Group 3: Mediterranean/polarized media systems	Italy, Portugal, Spain, Greece
Group 4: Post-Communist media systems	Bulgaria, Romania, Poland, Hungary, Czech Republic

The dependent variable used is Press freedom and can be extracted from two main sources. The first is the Press Freedom Index (RSF)¹, which ranks 180 countries and regions according to the level of freedom available to journalists. It is a snapshot of the media freedom situation based on an evaluation of pluralism, independence of the media, quality of legislative framework and safety of journalists in each country and region. It does not rank public policies even if governments obviously have a major impact on their country’s ranking. Nor is it an indicator of the quality of journalism in each country or region. Ever since the 2002 index, countries have been given scores ranging from 0 to 100, with 0 being the best possible score and 100 the worst. This makes the index more informative and makes it easier to compare one year to another. However, for better understanding of the data and for the purposes of this analysis, we reversed the index; therefore, in the analysis that follows, a higher number indicates higher press freedom, while a lower number suggests a lower one (worst possible score).

Another variable that can be used as the dependent one capturing press freedom is the Freedom House measurement², which assesses the real-world rights and freedoms enjoyed by individuals, rather than governments or government performance per se. It is produced each year by a team of in-house and external analysts from the academic, think tank, and human rights communities. They use a broad range of sources (including news articles, academic analyses, reports from nongovernmental organizations, individual professional contacts, and on-the-ground research) and score countries/territories based on the conditions and events within their borders. A country/territory is awarded 0–4 points for each of 10 political rights indicators and 15 civil liberties indicators, which take the form of questions; a score of 0 represents the smallest degree of freedom and 4 the greatest degree of freedom. The political rights questions are grouped into three subcategories: Electoral Process (3 questions), Political Pluralism and Participation (4), and Functioning of Government (3). The civil liberties questions are grouped into four subcategories: Freedom of Expression and Belief (4 questions), Associational and Organizational Rights (3), Rule of Law (4), and Personal Autonomy and Individual Rights (4). The highest overall score that can be awarded for political rights is 40. The highest overall score that can be awarded for civil liberties is 60. The combination of the overall score awarded for political rights and the overall score awarded for civil liberties, after being equally weighted, determines the status of Free, Partly Free, or Not Free. The total Political Rights and Civil Liberties scores are equally weighted in this calculation, leading to the possible ranges.

The two dependent variables differ with respect to their definition. However, a correlation test between the two yields that they are highly correlated (Table 2) especially in the years 2008–2019 (correlation higher than 0.8). A number closer to 1 or −1 yields strong correlations among the two variables. Also, all correlations appear to be significant (p-values less than 0.05). Using both as dependent variables will yield some differences in the results due to their different definitions. We choose for our analysis the RSF index since its definition and methodology is more appropriate for our study. (It has larger variation in scores and is more widely used in the literature).

Table 2. Annual correlations between the two press freedom indices.

2002	0.5165	2011	0.8572
2003	0.3302	2012	0.8513
2004	0.5633	2013	0.8735
2005	0.7956	2014	0.8669
2006	0.6653	2015	0.8987
2007	0.7076	2016	0.8955
2008	0.8060	2017	0.9024
2009	0.8284	2018	0.8650
2010	0.8539	2019	0.8675

The independent variables for the study include the macro-economic variables: GDP per capita growth (annual %), Gross domestic savings (% of GDP), Unemployment (% of total labor force), Inflation (consumer prices, annual %), General government debt (% of GDP), General government fiscal balance (revenues-expenditures, % of GDP), and Interest rates (long-term government bond yields, 10-year, and short-term policy interest rate). These variables were added to the study in line with relevant literature that suggests that they constitute the most notable variables that can sufficiently depict a country’s economic conditions and the basic indicators signaling the behavior of an economy, as earlier analyzed.

These variables are also directly linked to the financial institutions and financial markets within a country, they exert a powerful influence on economic development, poverty alleviation and economic stability and boost economic growth (Cihak et al. 2012; Berlinger et al. 2022).

The macro-economic independent variables derive from the World Development Indicators database³ published by the World Bank, the primary collection of development indicators, compiled from officially recognized international sources, which presents the most current and accurate global development data available, and includes national, regional and global estimates. The World Development Indicators is a compilation of high-quality and internationally comparable statistics about global development and the fight against poverty. The database contains 1,400 time series indicators for 217 economies and more than 40 country groups, with data for many indicators going back more than 50 years.

Analytically:

GDP per capita is gross domestic product divided by midyear population. GDP at purchaser prices is the sum of gross value added by all resident producers in the economy plus any product taxes and minus any subsidies not included in the value of the products. It is calculated without making deductions for depreciation of fabricated assets or for depletion and degradation of natural resources. It is a measure of economic activity and is also used as a proxy for the development of a country’s material standard of living, economic expansion, development and prosperity (Barro and Sala-i-Martin 2004). The variable ranges from 7,400–81,923 (all data). A country’s level of GDP per capita may be higher than others, but the growth of GDP per capita might remain stable, which would suggest that the country is not experiencing expansion of its GDP; while a country with lower levels might have an increasing rate of GDP, indicating higher economic growth. The growth rate in our dataset ranges between −9.1 and 25.2.

Gross domestic savings are calculated as GDP less final consumption expenditure (total consumption). The variable ranges from 8.3–57.6 as a % of GDP. Unemployment refers to the share of the labor force that is without work but available for and seeking employment. The unemployment rate ranges from 2% to 27.5%. Inflation, as measured by the consumer price index, reflects the annual percentage change in the cost to the average consumer of acquiring a basket of goods/services that may be fixed or changed at specified intervals (e.g., yearly). Inflation rate ranges from −4.5% to 22.5%.

The variable General government debt (% of GDP) was obtained from the Global Debt Database (GDD)⁴ published by the IMF in 2020. The GDD is the result of a multiyear investigative process that started with the October 2016 Fiscal Monitor. The dataset comprises total gross debt of the (private and public) non-financial sector for an unbalanced panel of 190 advanced economies, emerging market economies and low-income countries, dating back to 1950.⁵ The public debt series corresponds to gross debt and aims at covering all debt instruments owed by the national government. These include: (i) loans; (ii) debt securities; (iii) currency and deposits; (iv) insurance, pension, and standardized guarantee schemes; (v) other accounts payable; and (vi) and special drawing rights. The variable as a % of GDP ranges from 11.9–184.8.

General government fiscal balance (revenues-expenditures, % of GDP) is calculated as general government revenues minus general government expenditures. The data were obtained from EUROSTAT⁶ and the OECD Database.⁷ Revenues encompass taxes, net social contributions, grants and other revenues. Expenditures comprise intermediate consumption, compensation of employees, subsidies, property income (including interest spending), social benefits, other current expenditures (mainly current transfers) and capital expenditures (i.e., capital transfers and investments). The data range from −0.9–51.6 as a % of GDP.

Long-term government bond yields—10-year (interest rate) were obtained from EUROSTAT⁸ and the OECD Database.⁹ Long-term government bond yields are calculated as monthly averages (non-seasonally adjusted data). They refer to central government bond yields on the secondary market, gross of tax, with a residual maturity of around 10 years. In our sample, they range from 0.02–22.5. The short-term policy interest rate was also obtained from the OECD Database, Monetary and Financial statistics. According to the OECD (2021), short-term rates are usually either the three-month interbank offer rate attaching to loans given and taken amongst banks for any excess or shortage of liquidity over several months or the rate associated with Treasury bills, Certificates of Deposit, or comparable instruments, each of three-month maturity. In our sample, it ranges from 0.11–29.1.

Data for Corruption were obtained from the Corruption Perception Index (CPI).¹⁰ The CPI scores and ranks countries/territories based on how corrupt a country’s public sector is perceived to be by experts and business executives. It is a composite index, a combination of 13 surveys and assessments of corruption, collected by a variety of reputable institutions. The CPI is the most widely used indicator of corruption worldwide. A country score indicates the perceived level of public sector corruption on a scale of 0-100, where 0 means that a country is perceived as highly corrupt and 100 means that a country is perceived as very clean. For our analysis, we have reversed the index so that a higher number will indicate higher corruption. The reversed Index ranges from 3 to 74. All descriptive statistics are presented in Table A1 (see Appendix).

Additionally, in an effort to capture possible effects that may derive from the interaction between economic variables under investigation, corruption and socio-political conditions affecting freedom of information legislature, we use the following variables drawn from the V-DEM database¹¹: Liberal democracy index, Civil liberties index and Political civil liberties index.

The Liberal democracy index is a variable created from the question: To what extent is the ideal of liberal democracy achieved? The liberal model takes a “negative” view of political power insofar as it judges the quality of democracy by the limits placed on government. This is achieved by constitutionally protected civil liberties, strong rule of law, an independent judiciary, and effective checks and balances that, together, limit the exercise of executive power. To make this a measure of liberal democracy, the index also takes the level of electoral democracy into account. This variable is an interval, ranging from low to high (0-1) (Coppedge et al. 2016).

The Civil liberties index is obtained from the question: To what extent is civil liberty respected? Civil liberty is understood here as liberal freedom, where freedom is a property of individuals, and constituted by the absence of violence committed by government agents, the absence of constraints of private liberties and political liberties by the government. This index is an interval, running from low to high (0-1) and is formed by point estimates drawn from a Bayesian factor analysis model (Coppedge et al. 2020).

Political civil liberties index comes from the question: To what extent are political liberties respected? Political liberties are understood as freedom of expression and association. The index is based on indicators that reflect government repression; it is an interval, ranging from low to high (0-1), formed by point estimates drawn from a Bayesian factor analysis model including the following indicators: government censorship effort — media, harassment of journalists, media self-censorship, freedom of discussion for men and women, freedom of academic and cultural expression (see analysis in Pemstein et al. 2020).

In our analysis, we use the levels of the variables (shares-%) instead of growth rates since they are more appropriate with respect to the estimation and explanation of the results. When using growth rates, the variation of most variables is small and centered around zero, leading to small estimation coefficients. With respect to economic variables, one can use either levels or growth rates. The signs of the coefficients are not usually affected. In economic growth studies, the literature uses both levels and growth rates, and this does not seem to affect the results (Mankiw, Romer, and Weil 1992; Ketteni, Mamuneas, and Stengos 2007)

Based on simple correlation analysis (Table 3), we observe that not all independent variables are correlated with press freedom (stronger correlation if the correlation coefficient is closer to −1 or 1). Moderate positive effects are observed between press freedom and savings as well as the indices representing democracy and civil liberties. Moderate negative effects are observed between press freedom, unemployment, debt and the short-term policy interest rate. Weaker positive relationships are found between press freedom, inflation, government fiscal balance and the 10-year bond. Economic growth and corruption seem to present very low correlation coefficients.

Table 3. Correlations among independent variables and press freedom.

	Press Freedom
GDP per capita growth	−0.06
Gross domestic savings	0.30
Unemployment	−0.26
Inflation	0.12
General government debt	−0.36
General government fiscal balance	0.13
CPI score	−0.54
10-year bond yield	0.04
Short-term policy interest rate	0.21
Liberal democracy index	0.54
Political civil liberties index	0.53
Civil Liberties index	0.58

As a correlation analysis usually presents only an indication of an effect and is not very appropriate when testing relationships (Wooldridge 2010), we proceed with a multivariate analysis. The purpose of the analysis is to test whether economic conditions affect the level of press freedom in a country. The general estimation model is given by: $y=X^{\mathrm{\prime }}\text{β}+u$where y is the press freedom index, X is a matrix of the independent variables (GDP per capita growth, gross savings, unemployment rate, inflation rate, government debt, government fiscal balance, corruption index and the 10-year bond yield along with the short-term interest rate, the liberal democracy, civil liberties, and political civil liberties) and u is the error term.

Our study is based on a panel dataset, having both a cross sectional and a time series dimension in which we follow the same unit (country) across time. For the statistical analysis of panel data, we cannot assume that the observations are independently distributed across time (unobserved factors affecting the dependent variable in a specific year may or may not affect it in another) so a conventional OLS estimation may not be appropriate.

Two methods are used to estimate panel data: fixed effects and random effects estimation (Wooldridge 2010). Fixed effects estimation allows for arbitrary correlation between ${\mathit{a}}_{\mathit{i}}$ and the explanatory variables in any period. The fixed effects model controls for all time-invariant differences between the units, so the estimated coefficients of the fixed effects models cannot be biased because of omitted time-invariant characteristics. Endogeneity may be an issue also with respect to the relationship. One needs to test if endogeneity is present and if so, an IV (instrumental variables) method applied to panel data (either fixed effect or random) should be used (Wooldridge 2010).

The Hausman test can be used to choose between the two estimation methods, which tests the random effects model versus the fixed effects model. The Hausman is a chi-square test, and if the probability (significance) of the test is <0.05, then we can reject the random effects, and instead use the fixed effects model. If the probability is >0.05, then the random effects model is preferred.

Predictive causality is an important aspect of applied time series and panel data analysis. Recently, Juodis, Karavias, and Sarafidis (2021) developed a new method for testing. the null hypothesis of no Granger causality, which is valid in models with homogeneous. or heterogeneous coefficients. This test offers superior size and power performance to existing tests, which stems from the use of a pooled estimator that has a faster convergence rate. The test has two other useful properties: it can be used in multivariate systems, and it is effective against both homogeneous and heterogeneous alternatives.

Following Juodis, Karavias, and Sarafidis (2021), we consider the following data model: $y_{it}={\varphi }_{0i}+\sum _{p=1}^P{\varphi }_{pi}{y}_{it-p}+\sum _{p=1}^P{\text{β}}_{pi}{x}_{it-p}+{ϵ}_{it}$for $i=1\dots n$ (country units) and $t=1\dots T$ (time). The parameter ${\varphi }_{0i}$ denotes individual specific effects, ${ϵ}_{it}$ are errors, ${\varphi }_{pi}$ denotes the heterogeneous autoregressive coefficients and ${\text{β}}_{pi}$ are the heterogeneous feedback coefficients or Granger causality parameters.

The null hypothesis that $x_{it}$ does not Granger cause $y_{it}$ is: $H_0:{\text{β}}_{pi}=0\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{c}\mathrm{h}i\mathrm{a}\mathrm{n}\mathrm{d}p.$The alternative is: $H_1:{\text{β}}_{pi}\ne 0.$

Failure to reject the null implies that $x_{it}$ does not Granger cause $y_{it}$. The same applies when $x_{it}$ is a vector of regressors. The test offers the opportunity to check for causality under cross-sectional heteroscedasticity, to use the BIC criterion to select the number of lags that provides the best model and to test whether all independent variables jointly Granger cause the dependent variables, along with a univariate test when modeling the independent variables separately.

Before applying the test, we recall that it assumes that the variables are stationary. One should consider first a set of unit root stationary tests for all independent variables separately. If a variable is found to contain a unit root, differencing the variable will provide a stationary series.

The null and alternative hypothesis for the specific test is: $H_0:\mathrm{P}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{l}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{s}\mathrm{u}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{o}\mathrm{t}\mathrm{s}$ $H_1:\mathrm{P}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{l}\mathrm{s}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{a}\mathrm{r}\mathrm{y}$Rejecting the null (p-value < 0.05) suggests that the series are stationary.

Results

We begin our analysis using the total sample (all 18 countries from 2002 to 2019) and then by group. We first present the test results and then proceed to the estimation. Firstly, before estimating the model, we test whether our series are stationary using the unit root tests (if p-value of the test is less than 0.05, reject the null; panels are stationary). The results are presented in Table 4.

Table 4. Unit root tests.

Variable	p-value
GDP per capita growth	0.0000
Gross domestic savings	0.0033
Unemployment	0.000
Inflation	0.0000
General government debt	0.0002
General government fiscal balance	0.0000
CPI score—corruption	0.0004
Bond yield −10 year	0.3664
Short-term policy interest rate	0.0001
Liberal democracy index	0.9294
Political civil liberties index	0.9940
Civil liberties index	0.9802

We can observe that all series are stationary except the 10-year bond yield and the indices for democracy and civil liberties. Taking first differences of the series and testing again suggests that the new bond series $\left(D\mathrm{b}\mathrm{o}\mathrm{n}{\mathrm{d}}_{it}=\mathrm{b}\mathrm{o}\mathrm{n}{\mathrm{d}}_{it}-\mathrm{b}\mathrm{o}\mathrm{n}{\mathrm{d}}_{it-1}\right)$ is stationary (p-value of test  = 0.0000).¹² The variable used to capture the long-term interest rate for the rest of our analysis is the stationary first differences. Similarly, taking first differences of the liberal democracy, civil liberties, and political civil liberties indices, and testing again point to stationary new series (p-values  = 0.000), which are now used in the rest of our analysis.

We proceed with tests to check if the errors ${\mathit{u}}_{\mathit{i}t}$ are homoscedastic and serially uncorrelated and if not, corrections are made. First, the assumption of homoscedasticity (null hypothesis) is tested using a Wald test. If the probability of the test is <0.05, homoscedasticity is rejected and therefore we need to correct using heteroscedasticity robust standard errors. Secondly, the assumption that the residuals across entities are not correlated (null) is tested, using the Pesaran CD test. If probability is <0.05, that would suggest correlation and therefore a correction will be needed. The Wald test for homoscedasticity is rejected in all cases, so corrections are made to incorporate heteroscedasticity in our data (heteroscedastic robust standard errors). The error correlation test has probabilities larger than 0.05, therefore the errors are uncorrelated.

To establish Granger causality, we perform the Juodis, Karavias, and Sarafidis (2021) test. We first test for joint causality (all independent variables jointly Granger cause the dependent), adding heteroscedasticity as an option in our analysis. This test will also show us the optimal lags to be used in the model.

The test is: $H_0:\mathrm{S}\mathrm{e}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{c}\mathrm{o}\mathrm{v}\mathrm{a}\mathrm{r}\mathrm{i}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{s}\mathrm{d}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{t}\mathrm{G}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{r}-\mathrm{c}\mathrm{a}\mathrm{u}\mathrm{s}\mathrm{e}\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{f}\mathrm{r}\mathrm{e}\mathrm{e}\mathrm{d}\mathrm{o}\mathrm{m}$ $H_1:H_0\mathrm{i}\mathrm{s}\mathrm{v}\mathrm{i}\mathrm{o}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}$Rejecting the null (p-value < 0.05) suggests that all independent variables jointly Granger cause the dependent variable (press freedom). The p-value test results (for total sample and groups) are 0.0000, indicating that all variables jointly Granger cause press freedom and the optimal lag (best fit model) is one. The univariate tests also suggest that all independent variables separately Granger cause press freedom (p-values  = 0.0000).

The literature suggests that causality may also run from economic growth to press freedom (Nguyen et al. 2021). Correlation analysis (Table 5) indicates that press freedom and CPI are correlated while the correlations among GDP, CPI and press freedom are not significant.

Table 5. Correlations among Press Freedom, Growth and CPI.

	Press freedom	GDP growth	CPI score
Press freedom	1	−0.055	-.541
GDP growth	−0.055	1	0.061
CPI score	-.541	0.061	1

We have also tested whether press freedom Granger causes growth. The p-values of the tests vary based on the group use, but they all suggest press freedom does not Granger cause economic growth (see Appendix, Tables A2–A6). Press freedom, though, does seem to Granger cause corruption (see Appendix, Tables A2–A6). This is true but not in all groups.

Finally, we proceed with the use of the Hausman test to choose between the two methods. The results from the Hausman test suggest that we use the fixed effects estimation method in all groups and total sample (this is the preferred method shown from the test). Endogeneity is an issue in country regressions. To deal with this problem, one can use an IV-2SLS procedure applied for panel data, and the fixed effect model proposed by Hausman. The instruments used are lagged values of the independent variables. A test for endogeneity (Tables A2–A6) is provided in order to follow the appropriate estimation method. The endogeneity test implemented is defined as the difference of two Sargan-Hansen statistics: one for the equation with the smaller set of instruments, where the suspect regressor(s) are treated as endogenous, and one for the equation with the larger set of instruments, where the suspect regressors are treated as exogenous. Also, the estimated covariance matrix used guarantees a non-negative test statistic. Under the null hypothesis that the specified endogenous regressors can be treated as exogenous, the test statistic is distributed as chi-squared with degrees of freedom equal to the number of regressors tested.

The results suggest that our regressors appear to be exogenous; as such, an Instrumental variables method is not needed. To deal with causality issues, two methods are proposed. The IV- 2SLS regression using instruments or the Difference-in-Difference estimation (which is very similar to the fixed effect model used in our analysis). Since IV methods are widely used when endogeneity is present (not the case here) and since choosing appropriate instruments is a major issue in econometric studies (especially cases where instruments are weak or not appropriate), we deemed it more appropriate to proceed with the fixed effect model estimation.

Results from all specification tests for the total sample and groups of countries are presented in the Appendix (Tables A2–A6).

Having established the appropriate data model, we proceed with the estimation results. First, we present the results for the total sample and then by group (Tables 6–10). The optimal lag used for our independent variables is one (suggesting that all independent variables are lagged by one, $x_{it-1}$), heteroscedasticity is imposed and the 10-year bond along with the indices from the V-Dem dataset are in first differences.

The results for the total sample are presented in Table 6, in regards to H1. As the analysis shows, GDP per capita growth, savings, unemployment rates, inflation, general government debt, general government fiscal balance as well as levels of corruption, and the short-term policy interest rates seem to affect levels of press freedom in all countries examined.

Table 6. Estimation results (total sample).

	Coefficient	Robust Standard Error	t-test	P>|t|
GDP per capita growth	0.35	0.12	2.81	0.005
Gross domestic savings	−0.28	0.14	−1.98	0.049
Unemployment	0.37	0.17	2.21	0.028
Inflation	0.91	0.27	3.41	0.001
General government fiscal balance	−0.59	0.12	−5.00	0.000
General government debt	−0.18	0.03	−6.57	0.000
Short-term policy interest rate	0.72	0.24	3.02	0.003
CPI score—corruption	−0.31	0.08	−3.73	0.000
Bond yield—first difference	0.01	0.26	0.06	0.955
Liberal democracy index—first difference	3.16	16.58	0.19	0.849
Political civil liberties index—first difference	25.21	30.56	0.83	0.410
Civil liberties index—first difference	47.06	37.48	1.26	0.210
Constant	113.37	5.15	22.03	0.000
R-square	0.3384
F-test (joint significance)	0.0000

However, significant differences are identified among different media systems, verifying in this way H2. In democratic/corporatist media systems, only two of the independent variables examined are identified as significant factors that can affect press freedom: inflation and corruption (Table 7).

Table 7. Estimation results—Democratic/corporatist media systems (Group 1).

	Coefficient	Robust Standard Error	t-test	P>|t|
GDP per capita growth	−0.29	0.27	−1.07	0.290
Gross domestic savings	0.29	0.45	0.65	0.516
Unemployment	−0.41	0.67	−0.62	0.540
Inflation	1.35	0.64	2.10	0.037
General government fiscal balance	−0.16	0.43	−0.37	0.716
General government debt	0.09	0.08	1.08	0.286
Short-term policy interest rate	0.16	0.66	0.25	0.803
CPI score—corruption	−1.15	0.25	−4.65	0.000
Bond yield—first difference	−0.67	1.25	−0.54	0.594
Liberal democracy index—first difference	45.00	43.42	1.04	0.304
Political civil liberties index—first difference	3.30	85.68	0.04	0.969
Civil liberties index—first difference	134.78	88.05	1.53	0.131
Constant	87.32	10.47	8.34	0.000
R-square	0.3577
F-test (joint significance)	0.0000

In liberal media systems, unemployment rates, inflation, general government debt, and corruption are identified as the most important independent variables that can affect levels of press freedom (Table 8).

Table 8. Estimation results—Liberal media systems (Group 2).

	Coefficient	Robust Standard Error	t-test	P>|t|
GDP per capita growth	−0.08	0.18	−0.43	0.672
Gross domestic savings	0.06	0.17	0.37	0.716
Unemployment	2.34	0.41	5.66	0.000
Inflation	2.24	0.50	4.45	0.000
General government fiscal balance	−0.28	0.18	−1.55	0.127
General government debt	−0.31	0.04	−7.35	0.000
Short-term policy interest rate	−0.37	0.51	−0.72	0.474
CPI score—corruption	−0.51	0.25	−2.08	0.039
Bond yield—first difference	−0.07	0.59	−0.12	0.901
Liberal democracy index—first difference	−24.37	28.86	−0.84	0.403
Political civil liberties index—first difference	17.68	84.35	0.21	0.835
Civil liberties index—first difference	32.71	47.34	0.69	0.493
Constant	96.99	7.86	12.33	0.000
R-square	0.3067
F-test (joint significance)	0.0000

In Mediterranean/polarized media systems, in countries with severe economic problems especially after the Great Recession, the variables that seem to affect press freedom are GDP growth, gross domestic savings, government fiscal balance and debt, the 10-year bond yield along with the short-term interest rate (Table 9).

Table 9. Estimation results—Mediterranean/polarized media systems (Group 3).

	Coefficient	Robust Standard Error	t-test	P>|t|
GDP per capita growth	0.41	0.23	1.81	0.077
Gross domestic savings	0.97	0.40	2.42	0.019
Unemployment	0.04	0.18	0.24	0.814
Inflation	0.52	0.40	1.31	0.198
General government fiscal balance	−0.66	0.16	−4.07	0.000
General government debt	−0.11	0.04	−2.97	0.005
Short-term policy interest rate	1.35	0.48	2.83	0.007
CPI score—corruption	−0.14	0.14	−1.02	0.313
Bond yield—first difference	0.33	0.19	1.75	0.087
Liberal democracy index—first difference	3.84	22.56	0.17	0.866
Political civil liberties index—first difference	20.51	70.58	0.29	0.773
Civil liberties index—first difference	−2.96	52.59	−0.06	0.955
Constant	84.85	13.59	6.24	0.000
R-square	0.5694
F-test (joint significance)	0.0000

Finally, in post-Communist media systems four variables seem to affect levels of press freedom to a significant degree: GDP growth, gross domestic savings, inflation and the general government fiscal balance (Table 10).

Table 10. Estimation results—Post-Communist media systems (Group 4).

	Coefficient	Robust Standard Error	t-test	P>|t|
GDP per capita growth	0.81	0.26	3.08	0.003
Gross domestic savings	−1.39	0.31	−4.45	0.000
Unemployment	0.38	0.36	1.04	0.303
Inflation	1.28	0.43	2.95	0.004
General government fiscal balance	−2.14	0.37	−5.72	0.000
General government debt	0.11	0.12	0.86	0.392
Short-term policy interest rate	−0.47	0.40	−1.17	0.245
CPI score—corruption	−0.22	0.15	−1.49	0.142
Bond yield—first difference	0.42	0.82	0.51	0.615
Liberal democracy index—first difference	0.33	24.47	0.01	0.989
Political civil liberties index—first difference	16.55	43.29	0.38	0.703
Civil liberties index—first difference	60.21	64.96	0.93	0.358
Constant	166.76	15.67	10.64	0.000
R-square	0.3072
F-test (joint significance)	0.0000

Discussion

As regards the initial aim of the study, this work finds a reverse relationship between press freedom and the economy, besides the one already examined in recent years (e.g., Nguyen et al. 2021). Indeed, whereas economic conditions tend to affect the levels of press freedom in all media systems examined (H1), there are certain differences (H2), on which the econometric analysis attempts to shed some light.

In the Democratic/corporatist countries, inflation and corruption seem to play the most significant role. In the Mediterranean/polarized media systems, GDP per capita growth, government debt and fiscal balance, corruption and both interest rates (long and short run) seem to be the main variables that affect levels of press freedom. In the Liberal media systems, press freedom tends to be affected by unemployment, inflation, and government debt as well as corruption. Finally, in the post-Communist media systems, press freedom seems to be affected by GDP per capita growth, savings, inflation, and government fiscal balance.

Regarding the role of corruption, while deemed important for press freedom around the world (in line with Nam 2012; Camaj 2013), it does not seem to affect all media systems equally. Recently, Solis and Antenangeli (2017) showed that corruption is associated with greater levels of government attempts at censoring the media. This study argues that corruption tends to affect the levels of press freedom in all media systems except the post-Communist ones. We do not claim that corruption ceased to exist in post-Communist media systems, rather that these systems seem to be evolving in parallel with it, even in recent years. Therefore, corruption may indeed exist in these media systems (Bašná 2019) but does not seem to affect the levels of press freedom more or less than it did in previous decades. Interestingly, levels of corruption seem to be rising in Democratic/corporatist media systems (Maniou and Ketteni 2022), which seems to affect levels of press freedom, as censorship and attacks against media professionals have increased in recent years (Hiltunen 2017; Carlsson and Poyhtari 2017). Moreover, as far as the overall liberal and civil rights are concerned (including the legislation framework as well as freedom of expression and government suppression), this work argues that although there seems to exist a direct significant relationship for the overall sample of countries examined, more in-depth research is needed in the country-to-country level so as to reach more robust conclusions.

In examining the initial taxonomy of media systems (Hallin and Mancini 2004/ 2013) as regards the relationship between press freedom and economic conditions, this work finds significant differences among countries of the same type, which indicate a need to test more deeply to revise our existing knowledge of media systems (in line with Bruggemann et al. 2014 and Humprecht et al. 2022). For example, significant similarities in effects on press freedom can be seen among Democratic/corporatists, which implies that this category still remains solid. On the other hand, significant differences observed in Liberal media systems may show that this category has evolved and countries initially included need to be re-examined, as they tend to present very different effects.

Certain variations exhibited among countries of the same media system could be explained by each country’s economic status, especially after the Great Recession. For example, the USA and Ireland were among the first countries to be hit by the Great Recession after 2008 (Ireland signed an MoU with the EU and received financial help). Although both managed to recover their economic losses after 2010, their economic status is far from comparable to that of Canada, which may explain the differences in the economic variables affecting press freedom.

Moving towards the European South (Italy, Portugal, Spain, Greece), the situation becomes more complicated as these countries present several dissimilarities. Portugal is the most notable example, as negative effects from unemployment and inflation tend to mainly affect levels of press freedom as opposed to other countries of the region, such as Italy and Spain. In line with Bruggemann et al. (2014), we see that Portugal seems to be gradually presenting more similarities to Liberal media systems than to Mediterranean ones.

Finally, post-Communist countries (Bulgaria, Romania, Poland, Hungary and the Czech Republic), although sharing a common political past in which institutional controls over the media were normal governmental practice (Gross and Jakubowicz 2012), tend to present several differences. Scholars studying post-Communist media systems argue that several of them seem to present similar characteristics to the Mediterranean/polarized ones (e.g., Voltmer 2008), for example Romania in this study. However, due to their many dissimilarities (which are also evident in the way these countries perceive press freedom, i.e., Hungary and Bulgaria versus Poland), this work is in line with previous studies (e.g., Hallin and Mancini 2013) in the field, who argue that existing western models of media systems cannot be applied in the case of post-Communist media systems, which are better understood and studied as a separate, hybrid type (Dobek-Ostrowska 2015).

Conclusions

This work sought to examine two research hypotheses. Both hypotheses are verified as both the state of the economy within a country can affect the level of press freedom (H1), and the effect of economic conditions on the degree of press freedom seems to vary among different media systems (H2).

We argue that economic conditions may indeed impact press freedom in democratic countries of the western world but the extent to which this is happening may vary among different countries. In addition, variation seems to exist among the specific economic variables that affect press freedom. Whereas the type of media system plays a certain role in this variation at least for some categories (e.g., Democratic/corporatists), it is not the only factor that affects the degree of press freedom.

So far, to a certain degree, James Carville’s 1992 quote seems verified: it is, indeed, the economy that plays a vital role in the amount of press freedom every country seems to enjoy. As such, current debates regarding increases/decreases of press freedom levels cannot be examined without taking into account the prevailing economic conditions. This became evident after the 2020 pandemic crisis, when journalists around the world repeatedly reported press freedom violations (Palmer 2022; Papadopoulou and Maniou 2021). But our concluding argument is that it is not only the economy, but also other socio-political differences among different countries that need to be examined as regards press freedom. This study is focused on the effects of economic conditions on press freedom and, therefore cannot indicate conclusively whether other socio-political factors have a greater or lesser effect on press freedom than the economy. Among these, future research could consider the evolution of western media systems’ characteristics and the different levels of press freedom they present in line with revised taxonomies of Hallin and Mancini’s (2004) work. In addition, social-political characteristics that affect each country need to be examined in future studies to fully understand the relationship between press freedom and the overall status of the economy, such as legislature, and political and civil liberties.

As with most studies, this work presents certain limitations, which derive from the fact that although our research covers a period of 18 years, a higher number of cases (countries in this study) and data regarding independent variables would have strengthened our results. This limitation could be overcome with the inclusion of more countries in future research (beyond those examined in the initial theory of media systems). In addition, as the official data for the dependent variable (press freedom—RSF index) go back only as far as 2002, this limitation will be feasible to overcome in years to come and will advance our findings with a more in-depth country-to-country analysis so as to test variations among countries of the same group of media systems. As such, we consider this work as a first attempt to identify the relationship under study here with press freedom as the dependent variable and economic conditions as independent variables as opposed to previous studies in the field (e.g., Nguyen et al. 2021) that have adopted the inverse relationship. The initial goal was to expand our understanding in an emerging research area and provide evidence that press freedom is contingent on the conditions of the economy in western countries. If we were to expand the discussion of this relationship in other countries beyond the western world, we would have to take into consideration other socio-political parameters that affect these countries and may lead to opposing results of this model (e.g., in the case of China).

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by University of Cyprus, Start Up Research Grant Scheme 2019-21.

Notes

1 Press Freedom Index available at: https://rsf.org/en/ranking

2 https://freedomhouse.org/countries/freedom-world/scores

3 https://databank.worldbank.org/source/world-development-indicators

4 www.imf.org/en/Publications/WP/Issues/2018/05/14/Global-Debt-Database-Methodology-and-Sources-45838

5 For more details on the methodology and definitions, see Mbaye, Moreno Badia, and Chae (2018).

6 For European countries; https://ec.europa.eu/eurostat/data/database

7 https://stats.oecd.org/

8 https://ec.europa.eu/eurostat/data/database

9 https://stats.oecd.org/

10 https://www.transparency.org/en/cpi/2020/index/nzl

11 www.v-dem.net

12 This is true for the total sample and also for each group.

Appendix A:

Descriptive Statistics and Specification Tests.

Table A1. Descriptive statistics.

	Minimum	Maximum	Mean	Std. Deviation
Press freedom	50.00	95.20	74.50	9.40
GDP per capita growth (annual %)	−9.10	25.20	2.10	3.00
Gross domestic savings (% of GDP)	8.30	57.60	23.70	7.90
Unemployment, total (% of total labor force)	2.00	27.50	8.30	4.50
Inflation, consumer prices (annual %)	−4.50	22.50	2.10	2.30
General government debt (% of GDP)	11.90	184.80	66.40	34.60
General government fiscal balance (% of GDP)	−0.90	51.60	30.20	5.07
CPI score (0 = very clean—100 = highly corrupt)	3.00	74.00	33.40	18.70
10-year bond yield	0.02	22.50	4.06	2.35
Short-term policy interest rate	0.11	29.10	3.08	2.93
Liberal democracy index	0.39	0.89	0.77	0.10
Political civil liberties index	0.67	0.98	0.95	0.04
Civil liberties index	0.82	0.97	0.93	0.03

Table A2. Specification tests for all countries.

Tests:	Probability	Decision
Hausman Test H0: Random Effects Model H1: Fixed Effects Model	0.000	Use the fixed effects estimation method
Wald Test H0: Homoscedasticity H1: Heteroscedasticity	0.000	Obtain heteroscedasticity robust standard errors
Pesaran Test H0: Residuals are not correlated H1: Residuals are correlated	0.388	Not correlated
Granger—causality test -Press Freedom H0: Selected covariates do not Granger—cause H1: Null is violated	0.0000	Covariates Granger—cause press freedom
Granger causality test—GDP growth	0.5444	Press freedom does not Granger—cause GDP growth
Granger causality test -CPI	0.0000	Press freedom Granger—cause CPI
Endogeneity test H0: Covariates are exogenous H1: Null is violated	0.1206	Regressors are exogenous

Table A3. Specification tests—Democratic/corporatist media systems (Group 1).

Tests:	Probability	Decision
Hausman Test H0: Random Effects Model H1: Fixed Effects Model	0.0392	Use the fixed effects estimation method
Wald Test H0: Homoscedasticity H1: Heteroscedasticity	0.0202	Obtain heteroscedasticity robust standard errors
Pesaran Test H0: Residuals are not correlated H1: Residuals are correlated	0.592	Not correlated
Granger—causality test -Press Freedom H0: Selected covariates do not Granger—cause H1: Null is violated	0.0000	Covariates Granger—cause press freedom
Granger causality test—GDP growth	0.108	Press freedom does not Granger—cause GDP growth
Granger causality test -CPI	0.023	Press freedom Granger—cause CPI
Endogeneity test H0: Covariates are exogenous H1: Null is violated	0.2033	Regressors are exogenous

Table A4. Specification tests—Liberal media systems (Group 2).

Tests:	Probability	Decision
Hausman Test H0: Random Effects Model H1: Fixed Effects Model	0.0004	Use the fixed effects estimation method
Wald Test H0: Homoscedasticity H1: Heteroscedasticity	0.512	Homoscedasticity
Pesaran Test H0: Residuals are not correlated H1: Residuals are correlated	0.311	Not correlated
Granger—causality test -Press Freedom H0: Selected covariates do not Granger—cause H1: Null is violated	0.0000	Covariates Granger—cause press freedom
Granger causality test—GDP growth	0.569	Press freedom does not Granger—cause GDP growth
Granger causality test -CPI	0.145	Press freedom does not Granger—cause CPI
Endogeneity test H0: Covariates are exogenous H1: Null is violated	0.575	Regressors are exogenous

Table A5. Specification tests—Mediterranean/polarized media systems (Group 3).

Tests:	Probability	Decision
Hausman Test H0: Random Effects Model H1: Fixed Effects Model	0.000	Use the fixed effects estimation method
Wald Test H0: Homoscedasticity H1: Heteroscedasticity	0.000	Obtain heteroscedasticity robust standard errors
Pesaran Test H0: Residuals are not correlated H1: Residuals are correlated	0.463	Not correlated
Granger—causality test -Press Freedom H0: Selected covariates do not Granger—cause H1: Null is violated	0.000	Covariates Granger—cause press freedom
Granger causality test—GDP growth	0.409	Press freedom does not Granger—cause GDP growth
Granger causality test -CPI	0.509	Press freedom does not Granger—cause CPI
Endogeneity test H0: Covariates are exogenous H1: Null is violated	0.848	Regressors are exogenous

Table A6. Specification tests—Post-Communist media systems (Group 4).

Tests:	Probability	Decision
Hausman Test H0: Random Effects Model H1: Fixed Effects Model	0.000	Use the fixed effects estimation method
Wald Test H0: Homoscedasticity H1: Heteroscedasticity	0.014	Obtain heteroscedasticity robust standard errors
Pesaran Test H0: Residuals are not correlated H1: Residuals are correlated	0.383	Not correlated
Granger—causality test -Press Freedom H0: Selected covariates do not Granger—cause H1: Null is violated	0.000	Covariates Granger—cause press freedom
Granger causality test—GDP growth	0.650	Press freedom does not Granger—cause GDP growth
Granger causality test -CPI	0.000	Press freedom Granger—cause CPI
Endogeneity test H0: Covariates are exogenous H1: Null is violated	0.629	Regressors are exogenous