DEFINING MEDIA FREEDOM

In order to gather and measure data on media freedom, we must first define it. The United Nations Universal Declaration of Human Rights calls for freedom of expression and specifies that this includes the right “to seek, receive and impart information and ideas through any media and regardless of frontiers,” (United Nations 1948). Yet, though almost all countries have some constitutional provision for media freedom or freedom of expression, these provisions seldom guarantee these freedoms in practice (Breunig Reference Breunig1994). Thus, laws promising media freedom are not necessary an indication of it.

From Milton’s (1644) argument for an end to pre-publication censorship to the Committee to Protect Journalist’s (2011) “Attacks on the Press” report, there is far more clarity about the factors that constrain media freedom than there is regarding the criteria that establish it (Milton Reference Milton2010). For example, McQuail defines media freedom as “the right to publish without any prior censorship or license and without incurring penalties, within the limits of other legal obligations” (Reference McQuail2000, 146–7). He traces the roots of media freedom to the 18th century writings of Revolutionary Pamphleteer Thomas Paine and British Statesman Edmund Burke, who famously declared the press to be the “fourth estate.” Both Paine and Burke focused on the potential political power of journalism, which depends on the ability of the press to criticize government.

Our definition of media freedom returns to these roots and the primary reason that its defenders provide for justifying its necessity—the idea that a free press is able to hold those in power accountable. Following Van Belle (Reference Van Belle2000) we conceptualize media freedom as an environment in which journalists are able to safely criticize political and economic elites at both the national and local levels. Of course, the capability to criticize does not necessarily lead to a propensity to do so. One of the primary criticisms of news media in the United States is that they often fail to report critically about government action (see, e.g., Bennett, Lawrence and Livingston Reference Bennett, Lawrence and Livingston2007), yet compared with their counterparts in China, journalists in the United States are remarkably free to criticize the government, should they choose to do so. This simple definition focuses on the environment in which media function. Journalists’ capability to criticize powerful elites can be compromised by government censorship and commercial pressures, as well as influential actors, including organized crime, opposition parties and religious organizations (Czepek Reference Czepek2009). We propose that the combination of these factors shapes the media environment. If government exerts complete control over news media or fails to keep other actors from constraining watchdog journalism, then media are Not Free (cases of not free media in 2012 included Mexico, Singapore and Armenia). If media are subject to some pressures and controls, but remain capable of criticizing elites, then media are Imperfectly Free (in 2012 Italy, India and Uganda fell into this category). If media are generally free to engage in watchdog reporting, then media are Free (Cape Verde, Australia and Uruguay were classified as free in 2012). We clarify the criteria used to classify the level of media freedom in the next section.

MEASURING MEDIA FREEDOM

The Global Media Freedom Dataset is an updated version of a definition-driven data set first collected by Van Belle (Reference Van Belle2000) and expanded by Whitten-Woodring and Van Belle (Reference Whitten-Woodring and Van Belle2014). The media environment for each country is placed in one of the following categories:

• ▪ 1—Free: countries where criticism of government and officials is a common part of the political dialogue.

• ▪ 2—Imperfectly Free: countries where social, legal or economic costs related to the criticism of government or officials limits public criticism, but investigative journalism and criticism of major policy failing can and does occur.

• ▪ 3/4—Not Free: countries where it is not possible to safely criticize the government or officials, and media are either indirectly controlled (coded 3) or directly controlled (coded 4).Footnote ¹

• ▪ 0—No Media: countries where there is no effective national media.

Only the Congo (1960–1968) and Nepal (1948–1959) fall into the No Media category. There are also several codes for missing data (8=missing because political or social disruption make it impossible to code the country’s media; 9=insufficient data). The three basic categories of Free, Imperfectly Free and Not Free reflect the definition of media freedom as the ability to safely criticize the government. Although intuitively this definition suggests a binary measure where journalists either can or cannot criticize the government, in reality there are many cases where journalists can criticize government, but will pay some costs for doing so; hence, the need for the Imperfectly Free category. Thresholds determine the categories, and the coding scheme was designed around this conceptualization.

To operationalize this threshold-based definition, coders were given two critical questions. To distinguish between the most important divide, the Not Free and the Imperfectly Free categories, coders were asked, “Could a domestic news organization publish or broadcast the full story of a government scandal on the scale of Watergate?” If “no,” then the coding was Not Free. If “yes,” then the coders were asked, “Is the criticism of government, political leaders and economic elites sufficiently costless as to appear routine?” If “yes,” then the coding was Free. Otherwise, the coding was Imperfectly Free.

We then employed a multi-stage multi-coder process (at least two evaluated each country) to assess the media environment for each country-year case based on the critical questions.Footnote ² In most cases there was coder agreement for each country-year. Where there was not, two additional coders (including at least one of the principal investigators) became involved. A limited number of cases involved conflicting information and required extended investigation by these additional coders. In these cases the conclusions were vetted and discussed by a small group involving the principal investigators. Qualitative information documenting the reasons for the shifts in media freedom in each of the 196 countries included in this data set is available in the Historical Guide to World Media Freedom (Whitten-Woodring and Van Belle Reference Whitten-Woodring and Van Belle2014).

Our goal in gathering this data set is to provide an historic and consistent measure of media freedom. There are other measures of media freedom, but these cover only recent decades and focus primarily on constraints on media freedom.Footnote ³ Table 1 presents a comparison of these different data sets. Only the GMFD provides data going back to 1948. Additionally, the GMFD offers a simple consistent coding scheme defined by thresholds and based on a clear and simple definition of media freedom. The Freedom House and Reporters Without Borders indexes use coding schemes that have changed over time and are based primarily on identifying media restrictions rather than on a definition of media freedom. Both indexes identify the status of the news media for each country-year, but these statuses are determined by cut-offs in the scale rather than by thresholds. The IREX index is unique in that it focuses on media sustainability, but it includes a limited number of countries (80) for a limited number of years. Of these measures, Freedom House’s Freedom of the Press (FOP) index since 2001Footnote ⁴ provides the most inclusive coverage. We compared our codes for 2001 with those of Freedom House and found them to be highly correlated (0.864). Given this correlation and the detailed annual reports provided by Freedom House for all available countries since 2001, we decided to base much of our post-2001 data on the information in the Freedom House FOP reports. The FOP assigns each country-year a score from 0 to 100, with 100 being the most restricted, and categorizes each media environment as “Free” (those with a score of 0–30), “Partly Free” (31–60) or “Not Free” (61–100). While these categories for the most part correspond with our Free, Imperfectly Free and Not Free categories, the borders between the FOP categories are less distinct than those of the GMFD. In the FOP, the difference between countries with a score of 60 and 61 is slight, whereas in the GMFD the difference between countries that are Imperfectly Free and Not Free is the difference between being able to criticize the government and not being able to criticize the government. As a result in some cases, usually where the Freedom House scores are closest to the category borders, the FOP categories do not correspond with the GMFD categories. Thus, we have used the FOP data as a guide, but have also relied on historic accounts and our own coding criteria for the years since 2001.

Table 1 Comparison of Global Media Freedom Dataset and Other Data Sets

NGOs=non-governmental organizations.

The GMFD provides a categorical coding rather than interval scale; the difference between media coded Imperfectly Free and media coded Not Free is more substantial than the difference between those coded Free and Imperfectly Free. Also, the coders of both the original data set and the updated data set identified a notable bimodality to the data. Most states were clearly on one side or the other of the divide between the two functionally free categories and the Not Free category. Because of the bimodal nature of these data, when using media freedom as an independent variable, we typically collapse categories 1 and 2 to form the value “free media” and category 3/4 to form the value “not free media.” When using media freedom as a dependent variable (as we are here) we recommend using the categories Free, Imperfectly Free and Not Free because there are probably very different reasons that lead a country to have Imperfectly Free media rather than Not Free or Free media. For example, some resource-poor authoritarian governments might allow Imperfectly Free media as a means to keep track of lower level bureaucrats (Egorov, Guriev and Sonin Reference Egorov, Guriev and Sonin2009) or to gain legitimacy in the international arena (Whitten-Woodring Reference Whitten-Woodring2009). Imperfectly Free media in a non-democratic setting might also be an indicator that the regime does not have the capacity to control media; evidence of this can be seen in developing countries such as Tanzania, Libya (2011–2012) and the Maldives. In democratic settings, Imperfectly Free or Not Free media may be the result of third party actors attacking and threatening news media. A case in point is Mexico, which for years was a one-party state with Imperfectly Free media and is now a democracy with Not Free media, largely because of the drug cartels. Another example of this is the Philippines where media are Imperfectly Free, but attacks on journalists carried out by non-government actors largely go unpunished, including the killing of 32 journalists in 2009 in the Maguindanao massacre.

Interestingly, regime type does not always determine a state’s media system. Figure 1 depicts the distribution of Free, Imperfectly Free and Not Free media across a range of regime types, with regime type based on the Polity Score, which is a measure of institutional characteristics and ranges from −10 (most autocratic) to 10 (most democratic).Footnote ⁵ As expected, most cases of Not Free media appear in autocratic countries and most cases of Free media appear in democracies, yet there are cases of Not Free media in democracies, including Colombia (2000–2005), Portugal (1976–1994), and Burundi (2005–2012), and Imperfectly Free or Free media in non-democracies, including Panama (1948–1968), Senegal (1981–1999) and Thailand (2006–2009). To demonstrate how this data set can be used, we next present a preliminary analysis of the correlates of media freedom.

Fig. 1 Not always where it’s expected: kernel density plots of the dispersion of media freedom across regime types in 169 countries from 1948 to 2010

THE CORRELATES OF MEDIA FREEDOM

Although recent studies have investigated the effects of media freedom (Van Belle Reference Van Belle1997; Choi and James Reference Choi and James2006; Whitten-Woodring Reference Whitten-Woodring2009; Schoonvelde Reference Schoonvelde2014; Stein and Kellam Reference Stein and Kellam2014), few have considered the factors that might promote or discourage media freedom. In fact, much of the activism and policies concerning media freedom have been based on assumptions of what it is and how it can be fostered. This study returns to the initial empirical foundations of the subject and examines several proposed or assumed correlates of media freedom with the aim of identifying any necessary and/or sufficient conditions for media freedom.

Specifically, we begin by testing two hypotheses developed by Nixon (Reference Nixon1960, Reference Nixon1965):Footnote ⁶

Additionally, we test a new proposition:

This set of hypotheses is neither comprehensive nor exclusive; rather it is a starting point in investigating the correlates of media freedom.

To test these hypotheses, we used the GMFD and data from a variety of sources (these include several control variables—all are identified and summarized in Table 2), and conducted a series of analyses using multinomial logistic regression.Footnote ⁷

Table 2 Data and Sources

GDP=gross domestic product.

Overall, the results of our analyses indicate that previous media freedom and executive constraints consistently have statistically significant and positive effects on media freedom. Wealth (as measured by GDP/capita) and internet penetration generally have positive effects, whereas our control variables of civil conflict, international conflict and oil reserves generally have negative effects on media freedom. Additionally, there are some interesting differences in the effects of some of these variables in democracies versus non-democracies.

We first analyzed the effect of our primary independent variables across all countries for which data were available (149) from 1951 to 2011 (Model 1 in Table 3). We then added more recently available variables to these models, which shortened the time frame to 2000–2011 (Model 2 in Table 3). The coefficients reported show the effect of each independent variable on the likelihood of Imperfectly Free media over Not Free media, and Free media over Not Free media. Both models predict that, on average, countries with past media freedom and executive constraints are more likely to have Imperfectly Free media or Free media than Not Free media. When the effects of civil and international conflict are statistically significant, they are negative. In particular, Model 1 predicts that, holding other factors constant, countries experiencing civil conflict will be less likely to have Free media than Not Free media, and countries involved in international conflict will be less likely to have Imperfectly Free media than Not Free media. Both models predict that an increase in GDP/capita is associated with an increased likelihood of Free media over Not Free media.Footnote ⁸ Internet penetration had a statistically significant and positive effect and oil reserves had a statistically significant and negative effect on the likelihood of Free media over Not Free media, but neither internet penetration nor oil reserves had a significant effect on a country’s chances of having Imperfectly Free over Not Free media.

Table 3 The Correlates of Media Freedom

Note: these results are from multinomial logistic regressions with regional and year fixed effects and robust standard errors, clustered by country. Multinomial logistic regression reports the coefficients for each independent variable on each outcome of the dependent variable relative to the base outcome of the dependent variable. Here Not Free is the base outcome. The results show the effect of a one unit change in each independent variable on the likelihood that a country will have Imperfectly Free media compared with Not Free media (top rows of the table) and Free media compared with Not Free media (bottom rows of the table).

GDP=gross domestic product.

*p<0.05, **p<0.01, ***p<0.001.

Because we expected the factors that influence media freedom might vary depending on regime type, we conducted a set of analyses separating democracies from non-democracies. In general, these results (reported in the online appendix Results section) show that executive constraints have statistically significant and positive effects on media freedom, especially in non-democracies. In non-democracies, once we controlled for other factors, wealth did not seem to make a difference in media freedom, but in democracies wealth had a statistically significant and positive effect. Interestingly, civil conflict did not appear to influence the level of media freedom in non-democracies, but it did have a statistically significant and negative effect in democratic settings, whereas international conflict only had a statistically significant and negative effect in non-democratic settings.

Thus, we found that wealth is associated with increased media freedom, but only democracies; there is some support for Hypothesis 1. As hypothesized, executive constraints have statistically significant and positive effects on media freedom, especially in non-democracies. These effects are robust across different specifications, time frames and samples; Hypothesis 2 is supported. Hypothesis 3, the hypothesis that internet penetration is positively associated with media freedom is also supported.