1. Introduction
The benefits of economic freedom and good governance in terms of long-run growth and development are well documented (Acemoglu and Robinson, Reference Acemoglu and Robinson2006; Berggren, Reference Berggren2003; Hall and Lawson, Reference Hall and Lawson2014). Recent studies also suggest that more economic freedom is associated with peace, lower levels of crime, higher subjective well-being and more entrepreneurial activity (e.g. Bjørnskov, Reference Bjørnskov2015; Nyström, Reference Nyström2008; O'Reilly and Powell, Reference O'Reilly and Powell2015; Pitlik et al., Reference Pitlik, Redin, Rode, Gwartney, Lawson and Hall2015). However, the association between institutions and policies consistent with economic freedom and the structure of political institutions remains disputed (de Haan and Sturm, Reference De Haan and Sturm2003; Giavazzi and Tabellini, Reference Giavazzi and Tabellini2005; Pryor, Reference Pryor2010; Rode and Gwartney, Reference Rode and Gwartney2012).
The Hayek–Friedman hypothesis is central to these discussions. It states that societies with high levels of political freedom, usually interpreted to mean effective democratic rights to vote and run for election extended to all adults, and the possibility of influencing political decisions peacefully through other channels, must also maintain relatively high levels of economic freedom. Specifically, Hayek (Reference Hayek1944) and Friedman (Reference Friedman1962) argued that political freedom cannot be sustained over time if private markets become tightly regulated or directly controlled by the government. As Friedman claimed, a capitalist society with economically free citizens is a necessary (though not sufficient) prerequisite for stable democracy because competitive markets separate economic power from political power.
Conversely, in the opposite situation democracy becomes effectively restricted because the regime restricts freedom of expression and of the press. As Friedman (Reference Friedman1962: 23–4) argued, when government controls economic life, a “hypothetical supporter of capitalism would have to persuade a government factory making paper to sell to him, the government printing press to print his pamphlets, a government post office to distribute them among the people, a government agency to rent him a hall in which to talk, and so on.” Although substantially less precisely – Hayek continually argued that it was self-evident – Hayek (Reference Hayek1944) also saw economic freedom as political freedom.
Yet Hayek and Friedman each formulated a version decades ago and the hypothesis has since then been repeatedly criticized. The critique tends to focus either on the ground that neither writer clearly specified the mechanisms through which it works, or on the potentially weak empirical evidence for the hypothesis. The starting point of the present paper is that the central mechanism that requiring that limited economic freedom leads to a direct loss of free and fair elections – testing the hypothesis directly – is difficult to justify conceptually and may be too much to ask in the relatively short panels that are available today. Instead, I suggest that it is possible to test the main hypothesis by asking if economic freedom first affects the degree to which the regime – democratic or not – respects the freedom of the press.
The common theoretical mechanism alluded to in Hayek's and Friedman's contributions is indeed how extensively government control and a lack of economic freedom in most cases undermine press freedom, and thereby undermine citizens’ democratic rights by making government either independent of electoral control or able to control the electorate. The conceptual starting point is thus that freedom of the press is a precondition for having political institutions that can be democratically responsive to voters, and is a common ‘core’ of Hayek and Friedman's arguments. Focusing on press freedom, I first discuss in section 2 how economic freedom might theoretically affect press freedom. Section 3 presents the data, and section 4 illustrates the main developments and the background for the empirical strategy employed in section 5. The paper concludes with a discussion of the broader Hayek–Friedman hypothesis and the present findings.
2. The debate and mechanisms connecting economic freedom and press freedom
As noted in the introduction, the Hayek–Friedman hypothesis (HFH) has been repeatedly criticized, in particular since appearing in Friedman's seminal and highly influential 1962 book. The arguments against it in general take one of three directions: they either (1) argue that the understanding of political freedom in the HFH is too limited; (2) focus on the potentially weak empirical evidence for the hypothesis; or (3) criticize the hypothesis on the ground that neither Hayek nor Friedman clearly specified the mechanisms through which it works. I first briefly outline the main thrust of the debate after which I focus on the potential mechanisms connecting economic freedom to press freedom.
Defining political and economic freedom
The debate and some of its main arguments are muddied by conceptual disagreement when studies effectively argue against Hayek's – but not Friedman's – conception of political freedom (e.g. Zencey, Reference Zencey2012). The HFH is often discussed as a hypothesis about the association between economic freedom and democracy, but without clearly defining either concept. Friedman (Reference Friedman1962) defined a situation without economic freedom as one of centralized control of economic activities while Gwartney et al. (Reference Gwartney, Lawson and Hall2015: 1) define economic freedom as a state of society where ‘individuals are permitted to choose for themselves and engage in voluntary transactions as long as they do not harm the person or property of others’. While this is the operational definition in most research studying economic freedom, definitional standards are substantially more mixed in democracy studies.
The definition and operationalization of democracy remains a contested field, including everything from proponents of clean, minimalist definitions emphasizing the existence of free and fair elections with full suffrage that effectively allow the alternation of government power between different parties or interest (cf. Cheibub et al., Reference Cheibub, Gandhi and Vreeland2010) to much more maximalist definitions with strong normative components. These definitions include features of political freedom and human rights such as press and media freedom, freedom of assembly and freedom of speech, but some maximalist definitions also include a set of specific policies deemed to be normatively desirable as outcomes of democracy as well as the constitutionalization of socio-economic rights (Kerr, Reference Kerr2013; Morlino, Reference Morlino2004; Stilz, Reference Stilz2014).
Such conceptual disagreements are major problems when discussing the HFH as most critique against it tends to rest on a very particular maximalist definition of democracy. By applying a maximalist definition, this strand of literature effectively defines a normative and ideological ‘quality’ of democracy, which contrary to the HFH is inconsistent with economic freedom. Olssen (Reference Olssen2010: Chapter 3), for example, attacks a so-called ‘neoliberal’ conception of democracy as one that ‘avoids questions concerning the sharing of resources or the benefits of the world's wealth’. Yet the enforcement of socio-economic rights is, for example, logically inconsistent with the rule of law as it provides individual citizens with specific rights based on their belonging in particular groups or categories. Hahnel (Reference Hahnel2009), in a radical left-wing critique of the liberal market order and private property rights, instead claims that economic freedom actively undermines democracy, and advocates something she calls ‘participatory planning’. Common to all of these writers is a definition of democracy as a much more encompassing and maximalist concept than electoral democracy with free and fair elections and a full franchise. As such, their concept of a proper democracy is logically inconsistent with limited government.
Reich (Reference Reich2009) exemplifies another instantiation of the conceptual confusion that haunts political discussions surrounding this and related topics, where critics equate economic freedom with a neo-corporatist conception of society where government directly supports large industrial interests at the expense of average citizens, i.e. with a non-competitive version of capitalism. Such attacks on Friedman's version of the hypothesis claim that capitalism undermines democracy by creating inequality and giving undue power to large industrial interests. However, such a conception is also identical to what Zingales (Reference Zingales2012) calls ‘crony capitalism’ – a societal structure in which special interests are able to buy or otherwise procure policy and legislation that effectively limits or short-circuits competition and economic freedom. As such, Reich's critique of the HFH is remarkable only by employing a non-standard definition of the concept of economic freedom.
The mixed evidence
As such, some of the critical literature rests on conceptual definitions that are rather far from the way political and economic freedom is defined by Hayek and Friedman. The empirical evidence for the HFH remains more mixed, partly for similar reasons, and partly because Hayek in particular may not have thought of the hypothesis as testable in the same way as Friedman did.
Farrant and McPhail (Reference Farrant and McPhail2009) provide a representative example of the problem by summarizing the exchange between Hayek and Paul Samuelson, as well as some of the critique of Hayek's position. Hayek argued that he had issued a sincere warning about the likely path of economically unfree countries while Samuelson and several other commentators read the book as a direct prediction of a future to come (Alves and Meadowcroft, Reference Alves and Meadowcroft2014). The subsequent discussion, forming a background to the empirical studies, therefore came to revolve around the question of whether Hayek had described an ironclad law, providing a testable hypothesis stating that certain countries would tend towards autocracy, or merely engaged less formally in a normative political debate.
Samuelson (Reference Samuelson1970) nevertheless interpreted Hayek's argument as a directly testable and therefore falsifiable claim. More recent studies such as Farrant and McPhail (Reference Farrant and McPhail2009) and Alves and Meadowcroft (Reference Alves and Meadowcroft2014) have likewise emphasized mixed economies and the largest welfare states in northern Europe as counter-examples of Hayek's original hypothesis, as these are able to command about half of all income without degenerating into totalitarianism. Yet as shown by Bergh (Reference Bergh2014), the Nordic welfare states have only survived by continual reform that has left Denmark and Sweden in particular with large government sectors but otherwise with some of the least regulated economies and strongest private property rights in the world. His main point is that the protection and expansion of economic freedom – apart from government transfers and insurance – has proved necessary to finance the welfare state without the use of force.
More conceptually benign tests of the HFH – with a relatively minimalist definition including free elections, full participation and political competition, but not specific policy or institutional choices – suggest that the main thrust of the hypothesis tends to fit empirical tests. Lawson and Clark (Reference Lawson and Clark2010), for example, note that during the four decades in which measures of both freedoms have existed, no society without some minimum degree of economic freedom has succeeded in upholding democratic political freedom. While societies exist with high levels of economic freedom but little democracy or other aspects of political freedom – Singapore being the most obvious example – the opposite combination is extremely rare, suggesting that economic freedom is a necessary but not sufficient condition for sustained democracy. Similar studies associating the overall ideas with the hypothesis known as either ‘douce commerce’ or democratic peace also tend to find support (e.g. de Soysa and Vadlamannati, Reference De Soysa and Vadlamannati2013; Mousseau, Reference Mousseau2000). Yet the empirical literature remains mixed.
Mechanisms connecting economic freedom and press freedom
While empirical studies have successfully tested the main thrust of the hypothesis, the additional critique that the HFH is undertheorized remains. The most important strand of the HFH literature thus reflects the general informality and imprecision of the main argument. As such, several studies by researchers critical to the hypothesis have pointed out the imprecision, and attacked Hayek's arguments in particular for being inconsistent.
The starting point of this paper is that requiring limited economic freedom to lead to a direct and approximately immediate loss of democracy is both empirically impractical and theoretically uninformed, as it could involve a very large number of potential transmission mechanisms, which may prevent clear identification of any single mechanism. Instead, I approach the HFH from the common element in the two contributions by asking if economic freedom affects the degree to which a regime respects the freedom of the press.Footnote 1 The conceptual starting point is therefore that a sufficiently free press is arguably a necessary precondition for citizens, with the formal and effective freedom to vote, to be able to cast an informed vote (Leeson, Reference Leeson2008; Sen, Reference Sen1997; Strömberg, Reference Strömberg2015).
First, while the HFH is certainly undertheorized, this is partly a consequence of the general lack of consensus of how to define democracy. Focusing on one necessary element of political freedom – the freedom of the press – offers a situation in which it is easier to formulate more precise and more directly relevant theoretical mechanisms. It also begs two clear questions: who would be interested in limiting press freedom, and how would they go about doing it? At a basic level, two types of actors can have an underlying interest in limiting the freedom of the press: political actors and economic actors.
The political interests of limiting the information that is available to the general public are well described as a particular type of elite capture (e.g. Djankov et al., Reference Djankov, McLiesh, Nenova and Shleifer2003; Enikolopov and Petrova, Reference Enikolopov, Petrova, Anderson, Strömberg and Waldfogel2016). Any government has an incentive to stay in power, which is easier to achieve if the flow of information can be controlled so that any unfortunate news are either suppressed or only partially reported. With limited press freedom, government may also be able to place unchallenged information in the media that presents it in a favourable light. Having government control over the press thus allows government to avoid costly and potentially embarrassing social conflict, and to protect any political or economic perks that the general public may take issue with (Leeson, Reference Leeson2008). Following the introduction of special interest regulation and other policies that create economic perks for specific groups or the political elite, government may therefore have a particular interest in limiting press freedom in order to avoid opposition and a loss of perceived public legitimacy. In general, the more government does and the more ways it intervenes in society, the more likely is it that such situations will arise. Finally, limiting press freedom may also be a way to protect one's international reputation by limiting any potentially damaging information from being accessible abroad.
Economic special interests may also have an interest in limiting press freedom, arising out of a desire to prevent the public dissemination of information on either the existence or consequences of monopoly-like conditions, political connections, or unwanted consequences of economic activity such as pollution (Friedman, Reference Friedman1962). Similar interests may apply to religious organizations that often have particularly close political connections. Bjørnskov and Freytag (Reference Bjørnskov and Freytag2016) also hypothesize that a free press is likely to constitute a threat to organized crime and corrupt agents, which may subsequently threaten the life of journalists. In such cases, the media may be formally free in the sense of not being subject to legislative control, but will still be unfree since particular special interests – a Berlusconi in Italy, for example – are able to control the news flow. In such situations, government policies can create private incentives to limit press freedom through private means rather than through active lobbying for legislation. As such, limited press freedom may be either a reaction to legitimacy problems created by regulatory activity or a direct consequence of regulatory capture.
Subsequently, any mechanism connecting changes in economic freedom to intended as opposed to purely incidental changes in press freedom must logically rest on one of two types: either economic freedom that directly affects the ownership structure of the media or the de facto control that either government or large private and organizational interests have over the media (cf. Leeson and Coyne, Reference Leeson and Coyne2005). Regardless of the particular mechanism, reduced press freedom is therefore a logical consequence of elite capture that is also likely to be reflected in other regulatory policy and government support schemes.
Given that any agents are interested in limiting press freedom, the next question is the mechanism through which they do so. Strömberg (Reference Strömberg2015) emphasizes that media can serve three main political purposes: agenda setting, priming influence and framing. First, the most direct way of achieving the level of influence that allows a government or an economic interest necessary to control the agenda in public debate, for example, is to actively own the media. As such, Djankov et al. (Reference Djankov, McLiesh, Nenova and Shleifer2003) show empirically that government ownership of majority media is strongly associated with substantial losses of press freedom across most countries, as governments attempt to direct which information flows to the public. However, they also emphasize that the more homogeneous government-owned media's news coverage becomes, the better is the business case for alternative newspapers and other media. Despite even very concentrated ownership, it can therefore be difficult for private economic interests to control the media agenda (Petrova, Reference Petrova2011). Conversely, government may have an interest not only in controlling the domestic media but also in restricting access to foreign media. This almost always entails restricting market openness as information often flows through similar channels as trade and international market information. This type of mechanism is logically unavailable to private interests.
In total, there are a number of mechanisms through which the HFH could operate. Regulatory activity, the size and influence of the public sector, market openness and the general quality of public institutions all may affect press freedom in one or more ways. However, more recent thinking, in part inspired by Hayek's (Reference Hayek1960) study of constitutional political economy, provides a potentially important qualification, which I also take into account in the following. The main argument is that policies and institutional choices that affect economic freedom may be of relatively little importance if implemented in polities with weak veto institutions. Hayek (Reference Hayek1960) stressed the importance of checks and balances, and veto institutions that protect minority interests. Yet another consequence of having strong veto institutions that can block policy decisions is that they tend to make policy choices more persistent, as veto institutions also prevent policies from being changed (Keefer and Stasavage, Reference Keefer and Stasavage2003). Justesen (Reference Justesen2014) therefore argues along with Henisz (Reference Henisz2002) that the main effect of veto institutions is to lock in institutional changes such that institutional improvements are unlikely to be rolled back. Similarly, for the effects of policies changing economic freedom to affect de facto press freedom, sufficient political constraints must be in place. If not, journalists and other actors who take advantage of nominally improved press freedom may be hurt if policies change again. For policies to have full effect, central actors – in this case journalists and other actors within the media – must in other words expect that the policies are unlikely to be changed in the near future. As such, based on arguments that derive from Hayek's further work, the HFH is more likely to hold when veto institutions tend to lock in (both good and bad) policies and institutional choices, and unlikely to hold when relevant policies can be changed easily and thus becomes unstable. This and the more general HFH is tested in the following.
3. Data and estimation strategy
The main aim of this paper is to test the hypothesis that changes in economic freedom result in changes in press freedom such that more economic freedom will under very general conditions be associated with more press freedom. In order to test this hypothesis, I employ the Freedom House (2015) index of press freedom. The index is distributed between 0 and 100, with lower scores indicating more freedom; it is therefore a measure of the absence of freedom. The index is available since 1993 and decomposable since 2001 into sub-indices on law and regulations, the political environment and the economic environment. The index thus captures both political and legal limitations of press freedom as well as limitations created, for example by the ownership structure. In the following, I associate measures of economic freedom and a set of control variables between 1992 and 2010 with the Freedom House press freedom index in the following year, i.e. in the years 1993 to 2011. As all estimates in the following include country fixed effects, the estimates thus identify the association between within-country changes in economic freedom with changes in press freedom in subsequent years. This choice also means that all time-invariant features such as trade costs determined by geography, distance to central markets, historical characteristics and many other factors are effectively captured by the fixed effects.
Freedom House (2015: 25) takes its starting point in Article 19 of the Universal Declaration of Human Rights, which outlines the freedom of opinion, expression and press. The scoring rests on 23 methodology questions, seeking to capture ‘the varied ways in which pressure can be placed on the flow of information and the ability of print, broadcast, and internet-based media to operate freely and without fear of repercussions’ (ibid., 27). The overall index, which in principle ranges from 0 (perfect freedom) to 100 (no freedom) is composed of three sub-indices covering the legal, political and economic environments of the press. In the following, I reverse all indices of press freedom such that higher values intuitively are associated with more press freedom.
The first component of the index of press freedom encompasses the extent to which the freedom of the press is actively restricted by laws and regulations, and their enforcement in practice. The political environment covers ‘political control over the content of news media’, which includes editorial independence, censorship, information availability, and the de facto freedom of journalists to pursue their job. Finally, the economic environment contains a number of elements such as the industrial and ownership structure of the media, impediments of production, distribution and advertisement, and the potential effects of corruption on the media. The overall index thus includes assessments of a large variety of mechanisms that could affect the degree of press freedom, and has consequently been used in several recent empirical studies of the topic (e.g. Arrese, Reference Arrese2017; Dutta and Williamson, Reference Dutta and Williamson2016; Egorov et al., Reference Egorov, Guriev and Sonin2009; Stier, Reference Stier2015). It is, however, specifically focused on the media only and therefore does not overlap conceptually with the index of economic freedom except when general policy changes also come to affect institutional and policy changes particular to the press.
With respect to economic freedom, I employ the Heritage Foundation's (2015) Index of Economic Freedom (IEF), which is one of the two standard measures of the concept.Footnote 2 The index is a composite measure drawn together of four pillars: the rule of law, government size, regulatory efficiency and market openness. First, the rule of law is formed as the average of the protection of property rights and the freedom from corruption. Although based on subjective assessments, it is valid as it correlates highly with other measures of the rule of law, for example from Gwartney et al. (Reference Gwartney, Lawson and Hall2015). Second, government size also consists of two sub-indices, covering fiscal freedom, measured as the overall burden of all taxation as a percentage of GDP, and government spending, capturing the size of the public sector. Third, regulatory efficiency consists of business freedom, labour freedom and monetary freedom, measuring the absence of licensing and other directly limiting policies, hiring and firing regulations, and the existence of stable, predictable and non-inflationary monetary policy.Footnote 3 Finally, the indicator of market openness is formed from indices of trade openness measuring (1) average tariffs, the income from trade taxes, quotas and regulatory barriers to trade, (2) investment freedom, which captures transparent and equitable rules, and (3) an absence of restrictions on the movement of capital, and financial freedom, capturing transparent rules and the absence of government intervention in financial markets. These indices thus match the types of policies that are likely transmission channels of the HFH.
I also control for events that may change the political institutions, first of all the type of political regime of the country, as several recent studies find substantial differences (Møller and Skaaning, Reference Møller and Skaaning2013; Stier, Reference Stier2015). I follow the categorization in Cheibub et al. (Reference Cheibub, Gandhi and Vreeland2010), who identify six types: parliamentary, mixed and presidential democracies, and civilian, military and royal dictatorships. In the present context, it is important to emphasize that the conceptualization in Cheibub et al. (Reference Cheibub, Gandhi and Vreeland2010) is entirely minimalistic, such that the index only categorizes countries as democratic if citizens enjoy the effective right to vote and stand for parliament. The index thus does not contain any measures of press freedom per se. The inclusion of the regime type dummies also means that any potential immediate direct effects of economic freedom on electoral democratic development are controlled for.
As is standard, I add a set of economic control variables from the Penn World Tables, mark 7.1 (Heston et al., Reference Heston, Summers and Aten2012). The specification therefore includes the logarithms to population size and real GDP per capita, and a dummy indicator for negative year-on-year growth to account for effects of economic crises; all data are summarized in Table 1. In addition, I account for two types of coup d’état: successful and failed (cf. Wintrobe, Reference Wintrobe1998). The data in the following thus include 12 successful military and five successful civilian coups and 17 failed military and seven failed civilian coup attempts. Indicators for communism and membership in a commonwealth effectively capture the effects of transitions out of communism and the British Commonwealth.Footnote 4 As a final indicator of political institutions, I employ Henisz's (Reference Henisz2002) Political Constraints III index of veto player strength.
Table 1. Descriptive statistics
Finally, I use the political constraints measure in a set of additional analyses in which it is interacted with economic freedom, which provides tests of whether or not economic freedom exerts a stronger effect in conjunction with stricter political constraints, following the logic in Justesen (Reference Justesen2014). The interactions therefore capture how veto institutions change the marginal effectiveness of economic freedom.Footnote 5 All regressions in the following include a full set of annual and country fixed effects, such that joint international trends and time-invariant factors – culture, political history, colonial ties, etc. – are accounted for throughout. In separate tests, I rerun all regressions with subsamples that either include only democracies or exclude very poor countries, defined by Cheibub et al. (Reference Cheibub, Gandhi and Vreeland2010) and as those having a real GDP per capita below USD 4000 (reported in the appendix).
4. The simple picture
Figure 1, which plots the simple relation between economic freedom and press freedom in 2011, provides a first look at the data and first illustrates the strong correlation (r = 0.62) between the two types of freedom. Second, a comparison of the trend lines across observations with and without democracy indicates that the association is significantly stronger for democratic societies that are also generally freer, although the association is visible for autocracies. Yet the significant difference mainly rests on two outlier observations with high economic freedom and little press freedom: the oil state of Bahrain and the perennial outlier in international comparisons, Singapore.
Figure 1. Economic freedom and press freedom, 2010
Categorizing the 2,770 observations between 1993 and 2011 with full data provides a similar picture. Among the third of the observations with high economic freedom, 71% were also categorized as having free media by Freedom House, whereas 16% were partly free and 13% were considered unfree. In the third of observations with medium economic freedom, the corresponding shares were 27, 46 and 27% while the shares among observations with low economic freedom were 7, 36 and 57%. In other words, only 7% of observations had simultaneously low economic freedom and high press freedom, and the only countries in which Freedom House rated their media as free that at the same time had relatively low economic freedom for more than a few years are Suriname and Guyana; these countries were in the particular group in 11 and 14 years, respectively, between 1993 and 2011.
While the figure and the categorization exercise thus indicate the existence of a long-run equilibrium relation between economic freedom and press freedom, it could be spurious for several reasons. One could, for example, argue that overall economic development affects both or that both are consequences of the introduction of democratic political institutions. Two examples might serve to illustrate what the main data look like: the development in economic freedom and press freedom in Bangladesh and Venezuela from 1993 to 2010.
These recent examples tell similar stories to Figure 1. The Venezuelan attempt during the Chávez and Maduro regimes to create a socialist society with little economic freedom started early on to attack the political rights of its citizens. Following the development illustrates the dramatic policy changes that followed the 2000 elections, in which Chávez's socialist party won a supermajority and used it to get parliament to allow him to rule autocratically for 12 months by presidential decree. In the data used in the following, the reforms that Chávez started during this period resulted in a 20-point drop in the Heritage Foundation's assessment of Venezuelan market openness, accompanied by a contemporaneous 34-point deterioration in the Freedom House index of press freedom. The first 24-point drop occurred in the year in which a Chávez ruled without the parliament.
As a particular consequence of the much tighter regulation and government control of the economy, the Chávez regime were able to strike down hard on media that were critical of the regime. This was most clearly demonstrated when RCTV – Venezuela's oldest and most viewed network – lost its broadcasting licence in May 2007 after Chávez claimed that the station had incited the failed 2002 coup d’état. The decision was upheld by the Supreme Court, and all RCTV equipment was seized and made immediately available to the new government-owned TVes. The government also imposed severe fines on Globovision, on the grounds that it ‘generated civil unrest and violated a law for social responsibility from the media’ (Munoz, Reference Munoz2013).
Bangladesh provides another example, but one which the development is in the opposite direction. In Bangladesh, the democratically elected incumbent governments had suppressed the media since the late 1980s and the intensity of suppression and government control of the media increased during the 1990s. However, following the military coup in 2007, press freedom actually improved as Freedom House (2015) assesses both the legal and regulatory situation and the political environment as substantially less repressive than during the fragile democracy. These improvements were preceded by liberalizing reforms implemented by the new military government.
Similar developments took place in the Central African Republic and Fiji, following military coups in 2003 and 1999, respectively. Conversely, the effective democratizations in Indonesia, Mexico and Nigeria in 1999 and Peru in 2000 were preceded by substantial improvements in the FH index, i.e. by increasing press freedom. As such, these selected examples suggest that democracy is a far from sufficient condition for the freedom of the press to exist and be respected.
5. Empirical findings
These observations and the association illustrated by the figure are only indicative of a robust relation. In Table 2, I therefore present the results of estimating the relation between either the overall index of economic freedom or its four components and press freedom. I first of all find that more populous countries with weak veto institutions have less press freedom, as do countries with a communist regime or those experiencing a successful civilian coup d’état and those effectively leaving the British Commonwealth. Focusing only on the 114 countries with some experience with democracies in the sample, population size becomes negative while GDP per capita implies better press freedom.
Table 2. Main results
Note: * (**) [***] denote significance at p<0.01 (p<0.05) [p<0.10]. All regressions also include a constant term; numbers in parentheses are robust standard errors.
Turning to the main focus, the (lagged) IEF is significant in columns 1 and 3, indicating support for a press freedom version of the HFH. However, disaggregating the index in columns 2 and 4 shows that only particular components of economic freedom are significantly associated with press freedom: increased market openness is strongly associated with press freedom while an additional influence arises from the rule of law in democracies only. As such, the simple evidence in Table 2 suggests that liberalizing reforms of market policies and the judicial system are associated with subsequently improved press freedom.
Institutional conditions
Yet economic reforms can be short-lived and are always surrounded by some level of policy uncertainty. Justesen (Reference Justesen2014) therefore argues that property rights institutions in particular may only have long-run consequences when permanence is supported by strong veto institutions. In the additional results reported in Table 3, as well as in an online appendix, I therefore employ the specific areas of economic freedom in combination with an interaction with the Henisz measure of veto player strength.
Table 3. Results, single areas of press freedom
Note: * (**) [***] denote significance at p<0.01 (p<0.05) [p<0.10]. All regressions also include a constant term; numbers in parentheses are robust standard errors.
With interactions, neither the interaction term nor the separate estimate of the main variable can be interpreted on their own: Because the marginal effect is the un-interacted estimate plus the interaction estimate evaluated at any value of the interacting variable, one must calculate marginal effects for a number of values (Brambor et al., Reference Brambor, Clark and Golder2006). The bottom panel of the table therefore reports the conditional effects of each of the four components of economic freedom at the 10th, 25th, 75th and 90th deciles and the median of the distribution of veto player strength. These results thus evaluate the effects of economic freedom at low to high levels of veto institutions and make it possible to evaluate the changing effects of economic freedom depending on the degree to which veto institutions make them permanent.
In Table 3 and further robustness tests in an appendix, certain results persist in both the full sample and across democracies. While the rule of law seems to positively affect the law and regulation component of press freedom (i.e. it negatively affects the index), this result nevertheless fails when the poorest countries are excluded. Instead, the only robust result is that the rule of law positively affects the index capturing the economic environment when veto players have some strength. The government size component, on the other hand, seems to negatively affect press freedom, yet this result also disappears when the poorest are excluded.
For the regulatory component of economic freedom, results are mixed. When excluding the poorest countries (in the appendix), more regulatory freedom seems associated with worse legal regulation of the press when veto player strength is limited.Footnote 6 Conversely, it is associated with a better economic environment for press freedom when veto players have some strength. Finally, the strongest result is that in the economic environment, market openness is unconditionally beneficial for press freedom. However, given the significant correlation between three of the four components – government size is only weakly associated with the remaining three – collinearity is a potential worry. In particular, changes in components would arise if reforms tend to be implemented as packages that include changes to multiple components, which would impede the identification of component-specific effects.
Robustness tests
Yet while the findings above suggest that only certain elements of economic freedom are associated with press freedom, they may still either be statistically fragile or not generalize to most countries. In this final section, I therefore test their robustness.
As a test of the conditional findings, I first perform a ‘beauty contest’ in which I include interactions with two measures of economic freedom at a time. The beauty contests addresses the problem that the four different interactions are strongly correlated – interacting four correlated measures with the same variable logically yields heavily correlated interaction terms – which implies that it can be difficult in practice to separate the effects. With four indicators and three outcomes, the full set of combinations of two interactions yields 18 permutations. Requiring that any interacted effect must remain significant and relatively stable in terms of size, these 18 tests show clear support for three associations: (1) for law and regulation indicators of press freedom, the conditional results pertaining to regulatory freedom and market openness remain robust; (2) for indicators of the political environment, only the effects of government size remain stable; and (3) for indicators of the economic environment, the beauty contest supports the robustness of the effects of market openness and regulatory freedom.
As such, the occasionally significant associations with the rule of law turn out to be statistically fragile. Conversely, the remaining findings are robust to a set of additional tests. The main result – that the market openness component of economic freedom is responsible for the effects on press freedom – could for example be due to changes in trade flows that need not be associated with policy changes. Yet adding a set of extra variables – trade volumes as a percentage of GDP, Dreher's (Reference Dreher2006) KOF measure of globalization that compounds trade and investment flows, the KOF measure of specific trade and global investment policies, and the KOF measure of social globalization and international information flows – does not change this result. Even though economic globalization in the sense of both flows and policies appears significant, their inclusion only increases the size and significance of market openness.
Additionally, I test the robustness and generalizability of the main findings by excluding the tails of the distribution of press freedom, in order to ensure that the results are not driven by extreme press freedom, and excluding the tails of the distribution of economic freedom, such that results are not driven by extremely high or low freedom. I also exclude the ten potentially most influential countries in the sample, defined as those with the largest changes in press freedom. None of these tests change the main result, although the latter test reduces the size of the point estimate by about 30%. To the extent that such tests can inform, the findings therefore seem to generalize to most situations and countries, as argued by Hayek and Friedman.
6. Discussion
The Hayek–Friedman hypothesis originally stated that for democracy to be a stable political system, some level of market capitalism and economic freedom must exist. If not, democracy will gradually be undermined. Yet the hypothesis has been contested since Hayek's Reference Hayek1944 Road to Serfdom and returned with a vengeance when Friedman's seminal book Capitalism and Freedom raised the discussion in 1962. The discussion has revolved around two issues: (1) whether Friedman clearly specified the mechanisms through which it works, and (2) how strong the actual empirical evidence is for the hypothesis.
This paper argues that an intuitive way to test the HFH is to take Friedman's original arguments seriously and explore the association between economic freedom and press freedom. I argue that democracy has little meaning if voters do not have access to proper information that is not controlled by government or other large interests, i.e. if society enjoys press freedom (cf. Leeson, Reference Leeson2008; Sen, Reference Sen1997), and Friedman himself stressed press freedom as a mechanism through which his hypothesis worked. Focusing on press freedom thereby offers a way to formulate more precise hypotheses of the multitude of transmission mechanisms that could connect economic freedom and press freedom.
Testing the association between the overall Heritage Foundation IEF as well as its four constitutive components and the Freedom House index of press freedom shows that the association is statistically strong. In particular, measures of market openness – composed of the freedom to trade and invest internationally and the existence of transparent and equitable rules regulating investment and financial markets – and regulatory freedom – defined as the absence of licensing and other directly limiting policies, hiring and firing regulations, and the existence of stable, predictable and non-inflationary monetary policy – positively affect press freedom. However, the results also suggest that economic freedom affects press freedom more when veto institutions are strong such that reforms can be secured in the future by veto players. As such, the estimates provide statistically strong and robust evidence for the HFH when formulated as a theory of press freedom.
An important question nevertheless is whether the results are also of political and social significance, which the estimates suggest they are. The clearest example is liberalizing market openness by one standard deviation (approximately 12 points), which results on average in an improvement of press freedom of six points. This change corresponds to 40% of a standard deviation or a full within-country standard deviation; the corresponding numbers in the democratic subsample are 60 and 154%. Only the events of abandoning a communist dictatorship or experiencing a successful civilian coup that leads to a dictatorship are associated with similar or larger changes in press freedom.
The findings shed new light on at least one standard critique of the HFH: the sceptical argument that economic freedom in some way undermines the quality of democracy. While opponents of the hypothesis tend to define effective democracy as including redistributive policy and substantial market regulations, the present findings rest on a definition that includes the combination of free and fair elections and a potentially informed electorate. In his famous treatment of the association between political institutions and famine, Sen (Reference Sen1997: 7–8), for example – who is far from being a proponent of Hayek and Friedman – emphasizes that ‘in the terrible history of famines in the world, no substantial famine has ever occurred in any independent and democratic country with a relatively free press (added emphasis)’. As such, Sen puts particular importance on free access to information, and not simply the right to vote.
The present findings thereby suggest that the HFH holds for a slightly broader definition of democracy, although not one that includes specific interventionist policies. On the contrary, the results in this paper emphasize the value of policies and institutions of economic freedom in solving political problems associated with asymmetric information. As such, economic freedom not only makes countries richer, but may also have an intrinsic social and political value.
Acknowledgements
I gratefully acknowledge financial support from the Jan Wallander and Tom Hedelius Foundation. I am grateful to Niclas Berggren and two anonymous reviewers for encouragement and helpful comments on early drafts of the paper. Needless to say, all errors are entirely mine.
