Introduction

The diffusion of constitutional rights at a global scale has recently drawn heightened attention by comparative constitutional researchers,Footnote 1 not least thanks to several data coding efforts of constitutional provisions around the world. However, although the available studies compare up to 108 different rights and freedom provisions across national constitutions, academic freedom has not yet been among them.Footnote 2

One reason for this omission could be the fact that academic freedom, as a constitutional norm, is not clearly delimited. Academic freedom as a term is not explicitly defined in international law, but it is generally considered a professional freedom of those who engage in scientific research or higher education teaching and who are affiliated with a higher education or research institution. At a minimum, it protects their freedom to research, teach, disseminate their research, and exchange with other academics (cf. Altbach 2001; Beaud 2020; Finkin and Post 2009; Scott 2017). While the autonomy of higher education institutions is conceptually not a direct part of academic freedom, it is often considered a functional condition to its protection or otherwise connected through “indissoluble ties” (cf. Roberts Lyer et al. 2022: 9–25; Thorens 2006: 108). Within constitutional texts, norms related to academic freedom can therefore take various different shapes. Relevant provisions may explicitly include guarantees of “academic freedom”, or they may establish the freedom of “science”, of “higher education research”, of “higher education teaching”, the “autonomy of universities”, or similar wordings.

Although this diversity of provisions may partly account for the previous neglect of academic freedom in large-N constitutional studies, the International Covenant on Economic, Social and Cultural Rights (ICESCR)—one of the central rights catalogues that comparative constitutional researchers draw upon—does codify the “freedom indispensable for scientific research” as part of the right to science (Art. 15.3). Still, the freedom of science does also not appear in the relevant datasets. The omission of those concepts is unfortunate both from a normative and a substantive viewpoint. Normatively speaking, the protection of academic freedom as a fundamental value that promotes freedom of scientific research and teaching and the free exchange and expansion of human knowledge is a central element of democratic societies (Scott 2019: 14) and should therefore figure in analyses of fundamental rights protections. Substantively, the consideration of the global diffusion of academic freedom is relevant for comparative constitutionalist scholars because its patterns deviate from those of other rights and freedoms and may hold interesting lessons for the international diffusion of constitutional rights.

The Academic Freedom in Constitutions (AFC) dataset (Spannagel 2023) introduced in this article allows to fill this gap, providing new and comprehensive time series data covering 203 countries over a time span of more than 230 years (1789–2022). It details for each country and each year whether any reference to academic freedom is included in the text of the national constitution in force at the time and distinguishes three different reference types: (a) “academic freedom” specifically, (b) other phrases relating to (individual) academic freedom, such as the freedom of science, higher education teaching or research, and (c) the autonomy of higher education institutions. These distinctions aim to do justice—at least at a basic level—to the complex constitutional concept of academic freedom and allow to trace different reference types and genealogies of academic freedom provisions across regions and time.

For scholars studying constitutional design and norm diffusion, the dataset will be most useful for the above-mentioned comparisons with other constitutional rights and freedoms. But the data can also benefit scholars researching academic freedom specifically, as it provides systematic data for a better descriptive understanding of the constitutional codification status of academic freedom around the world and over time; and it allows to easily identify relevant cases for further qualitative study, for example to inform normative debates over academic freedom definitions, to explore differences in the legislative interpretations and jurisprudence on academic freedom, or to investigate drivers of changes in its constitutional protection over time.

This article will first explain how the new AFC dataset relates to and improves on existing datasets of constitutional academic freedom provisions. Then I give some details on how the data were collected and classified, as well as their temporal and geographic coverage. In the final section, the article discusses the dataset’s significance and possible applications in different research areas. In view of comparative constitutional research specifically, it provides some preliminary analysis of how and why the diffusion of academic freedom provisions may be different from related constitutional rights and freedoms, such as the freedom of expression or the independence of the judiciary. A major point seems not only its relatively late emergence as a constitutional norm in the first place, but also its close ties to the development of higher education, which impeded a widespread diffusion during critical periods of constitution-making. Even so, the impetus of individual norm-setting countries and regions has meant that it took hold in many constitutions across the world.

Comparison to existing datasets

The AFC data collection effort is similar to and inspired by the Comparative Constitutions Project's (CCP) coding of provisions in national constitutions, which actually includes a variable on academic freedom guarantees (Elkins and Ginsburg 2022). However, CCP's acfree variable has severe shortcomings both with regard to variable definitions and coding consistency. Moreover, though CCP is continuously expanding its coverage, it is currently still limited to only about 58% of all country-years in which constitutions were in force in independent states, or 79% when imputing data within constitutional systems (ibid: 6). According to CCP’s own assessment, the reliability of the acfree variable has been lower than the average reliability across their variables (cf. Spannagel and Kinzelbach 2022: n19), suggesting that the lack of definitional clarity in the codebook generated an unusually high discrepancy among coders who interpreted “academic freedom” in conflicting ways. Specifically, I found that problems relate to false positives arising from the frequent inclusion of provisions such as the “freedom of education” or “freedom of teaching” without any reference to higher education. On the other hand, phrases like the “freedom of scientific research” or “science is free” are inconsistently coded in the CCP dataset. For example, nearly identical provisions in the German constitutions are coded as “yes” in 1848 and 1949, but coded as “no” for 1919. Provisions on the autonomy of universities were generally not taken into account.

When comparing the new AFC data to CCP's acfree variable, we can see that the corrections introduced by the AFC’s coding protocol—and its application by a coder with expertise on academic freedom topics—are extensive: While 10,626 country-years were originally correctly classified, 1784 previously coded as “yes” are now coded as “no” (about 84% of which stem from “freedom of teaching” provisions), and 1239 country-years previously coded as “no” are now coded as “yes”. A total of 4116 country-years missing from CCP coding (or classified as “other”) are now coded for the first time in the AFC dataset.Footnote 3 Moreover, CCP’s coding does not make any distinctions between provision types, thus not allowing any in-depth investigations of their constitutional lineages.

A derivative version of the CCP variable used to be included in the Varieties of Democracy (V-Dem) dataset as variable v2caprotac for its releases v10-v12. As of v13, the AFC dataset replaces CCP as main source for v2caprotac (Coppedge et al. 2023).

Data collection

The corpus of constitutional texts reviewed

The AFC was created on the basis of a manual analysis of nearly a thousand constitutional texts. To build the corpus of texts to be reviewed, in a first step all countries that (potentially) ever had academic freedom provisions in their constitutions were identified based on: (a) whether the country had been identified by CCP as including academic freedom provisions at any point in time (94 countries), and (b) for all other 109 countries, whether their latest constitution included or includes any such provision (33 additional countries). The 127 thus identified countries were retained for detailed analysis. The 76 remaining countries were coded as not having constitutional academic freedom guarantees.Footnote 4

For the 127 countries under closer review, all new constitutions between independence and today were analysed (for a detailed collection of sources used, see the codebook at Spannagel 2023). I largely relied on CCP's structural variable of constitutional events to identify when new constitutions were introduced (Elkins and Ginsburg 2022). In addition, many in-between versions of constitutions altered by amendments were considered. Since English translations do often not reproduce the content faithfully at the required level of detail, they were compared with original texts wherever possible.

Classification of references

Whenever a constitution includes a relevant reference on academic freedom or institutional autonomy, AFC groups it into one of three main categories, which correspond to the following three variables in the dataset:

  • Academic: provisions on “academic freedom” (matching the exact phrase), including “libertad de cátedra” in Spanish,Footnote 5

  • Science: provisions on “freedom of science” (in a broader sense, also including phrases such as the freedom of research, of scientific investigations, of scientific creativity, of higher education teaching, of intellectual creation, etc.), and

  • Autonomy: provisions on “university autonomy” (in a broader sense, also including phrases like the autonomy, self-governance, independence of higher education institutions; that “institutions of science are free”, etc.).

The summary variable any_acfree is coded “1” if any of the above provisions is present in the constitution. In addition, for country-years with no such reference, a separate variable is coded as

  • Teaching_only: provisions on “freedom of teaching” or “freedom of education” without specified link to higher education (also including phrases like the freedom of learning, of acquiring and imparting knowledge).

This variable was included because many countries, especially in Latin America, had such provisions long before they included more specific guarantees relating to the higher education sector. The teaching_only variable allows to trace these developments over time.

Geographic and temporal coverage

For reasons of scope and availability, the AFC dataset is limited to 203 independent countries, thus excluding a number of historic countries and autonomous territories. In sum, the dataset includes the 193 current UN member states, Kosovo, Taiwan, and the following historic entities: Austria-Hungary, Czechoslovakia, the GDR, Korea, Serbia and Montenegro, South Vietnam, South Yemen, Yugoslavia. For each country unit’s coding over time, the AFC relies largely on CCP’s way of coding constitutional periods and categorization of constitutional events.Footnote 6 Coding for each country starts with the year of the first constitution’s adoption. The starting date of the dataset at 1789 defers to CCP’s choice as “the effective date of the United States constitution, the widely reputed ‘first’ document of its kind” (Elkins, Ginsburg and Melton, n.d.). Where several constitutions were in force in a given country-year, the coding reflects the latest one that was in force. Constitutions that were suspended in a given year are coded as “1” under suspended and “0” on any of the provisions.

Insights from and applications of the dataset

The dataset shows that there are 116 countries that have had academic freedom provisions in their constitution at some point over the past 233 years. 101 such provisions were still in place in 2022. The first-ever academic freedom provision was one guaranteeing the freedom of science in the (short-lived) German constitution of 1848. It was later reintroduced in the Weimar constitution of 1919, whose wording influenced many other constitutions. One of them was the Yugoslavian constitution of 1921, whose later adjusted formulation of “scientific and artistic creativity shall be unrestricted” in 1963 came to shape academic freedom provisions in socialist-inspired constitutions across the world. University autonomy as a constitutional norm first appears after the First World War, almost simultaneously in Uruguay (1918), Finland (1919), Estonia (1920), and Georgia (1921). The specific term of “academic freedom”, on the other hand, is comparatively rare in constitutional texts. Its Spanish version “libertad de cátedra”Footnote 7 is first enshrined in Spain’s 1931 constitution and likely travelled to Latin American constitutions from there. Outside the hispanophone world, the term notably appeared in the Filipino and Japanese constitutions of the 1940s, and then in the Liberian constitution of 1986, possibly pointing to US influences, where “academic freedom” (though not constitutionally protected) was already well established as a term and concept at the time (Barendt 2010: 161–201).

Figure 1 shows a map of academic freedom provisions as of 2022, for simplicity only differentiating between (1) “academic freedom”/freedom of science provisions and (2) university autonomy, as well as (3) where they overlap. Still only 52% of constitutions in force today have any such provision, though they are geographically spread out. Provisions on university autonomy are dominant in Latin America and also widely present in Eastern Europe.

Fig. 1
figure 1

Source: AFC

Academic freedom provisions in constitutions worldwide 2022.

Full size image

The new data are significant for both constitutional and academic freedom researchers in various ways. One potential use case of the data becomes clear when comparing them to recent efforts of understanding the freedom of science or academic freedom as constitutional rights, such as in Romano and Boggio (2020) or Vrielink et al. (2011). Both articles attempt to provide an overview of the geographic scope of such constitutional guarantees, but neither do so comprehensively. For instance, Romano and Boggio, in their entry for the Max Planck Encyclopedia of Comparative Constitutional Law, claim that “in the Americas, academic freedom and freedom of scientific research are guaranteed in five constitutions”, namely Peru, Ecuador, Brazil, Suriname, and Nicaragua. However, the AFC datasets identify ten additional countries (Colombia, Costa Rica, Cuba, Dominican Republic, El Salvador, Honduras, Mexico, Panama, Paraguay, and Venezuela) as having such provisions in that region—not counting those where only university autonomy is guaranteed (Uruguay, Bolivia, Haiti, Argentina). While the authors’ definitions may depart from the approach taken by AFC, this does not account for the wide discrepancy, as many of the omitted constitutions have similar provisions as the ones included by the authors. Either way, taking the AFC as a starting point can significantly facilitate the identification and evaluation of potentially relevant provisions.

Secondly, the AFC data can contribute to a better and more systematic understanding of the historical spread of constitutional academic freedom provisions. One author erroneously claimed that it was “only long after the second world war” that the freedom of science “occasionally appeared” as a constitutional right independent of the freedom of opinion in countries outside Germany (cf. Gärditz 2018: 16, own translation). In contrast, the AFC data show that many other states addressed the issue in their constitutions before the second world war, including Austria, Greece, and Spain in Europe; Brazil, Cuba, El Salvador, Nicaragua, and Peru in Latin America; Iran in the Middle East; and Estonia, Czechoslovakia, Georgia, Poland, and Yugoslavia in Eastern Europe and Central Asia (plus others that included the autonomy of universities). While not all of these countries specifically protected the “freedom of science” (some guaranteed academic freedom or the freedom of scientific research), many did. Having systematic data of such provisions over time can thus help to rectify accounts of the term’s constitutional history.

As to its global spread over time, the AFC data show that academic freedom (in a broad sense) started to be more widely represented in constitutions after the end of the second world war (see Fig. 2). By 1950, one-third of constitutions in force included at least one of the three provision types distinguished here. Over the following decades, however, the rate at which newly independent countries adopted constitutions outpaced that of newly adopted academic freedom provisions—the percentage among those in force fell to 23% in some years. Then, in the decade after the end of the Cold War, the vast majority of newly adopted constitutions contained such a clause—bringing the global percentage to over 40% in the mid-1990s and to just over 50% today, still remaining at a remarkably low level compared to many other constitutional rights (Goderis and Versteeg 2014; see also Fig. 2).

Fig. 2
figure 2

Source: AFC and CCP (Elkins and Ginsburg 2022)

Share of constitutions in force containing academic freedom compared to other provisions.

Full size image

On the basis of such a systematic analysis of certain rights provisions over time and space, a third application of the AFC dataset is its contribution to the comparative constitutions literature. The constitutional academic freedom data can, for instance, inform discussions on the extent and drivers of global constitutional convergence (cf. Cope et al. 2019: 169). Elkins, Ginsburg and Simmons, for example, observe “a qualified convergence on the content of rights”, noting that “nearly every single right has increased in prevalence since its introduction, but very few are close to universal” (2013: 61). Goderis and Versteeg find various mechanisms of constitutional rights diffusion, in particular that countries often learn from their peers or are influenced by aid donors or former colonizers (2014: 28). The omission of academic freedom provisions from these studies to date does certainly not invalidate their findings, nor would its inclusion change the broader patterns of constitutional (non-)convergence. However, there are two elements that make the diffusion of academic freedom provisions particularly interesting and which may distinguish them from other constitutional rights.

Firstly, academic freedom, with its close links to higher education, is dependent on the development of an institution external to the political system. While literally every nation in the world today hosts some kind of higher education institution on its territory,Footnote 8 this was not the case for a long period of time—only after the second world war did higher education expand at a global level (Schofer and Meyer 2005). This means that the issue of academic freedom was not salient at the time when many countries adopted their first constitutions, even well into the second half of the twentieth century. Combined with the fact that academic freedom was not among the first-generation constitutional rights, accounting for its absence in many influential constitutional texts, this impeded widespread diffusion especially at critical moments of constitution-making, such as decolonization. In fact, the AFC dataset indicates that most postcolonial independence constitutions did not include an academic freedom provision. The same was not true for functionally at least somewhat similar provisions such as the freedom of the judiciary, whose coverage grew with those independence constitutions (see Fig. 2). Moreover, academic freedom or the freedom of science was not explicitly included in the Universal Declaration of Human RightsFootnote 9—and omission from this document has caused even some previously popular rights such as the freedom of the press to fall “out of fashion” (Elkins et al. 2013: 81; see also Fig. 2). Seen in the light of these inhibiting factors, it seems almost surprising that academic freedom norms have still reached a coverage of more than half of the world’s constitutions in recent years. Patterns in the AFC data show that this recovery was mainly driven by serial adoptions in transitional constitutions of the 1990s in Eastern Europe and Central Asia as well as in Africa, many of which share a similar origin (see below).

Secondly, the examination of the different provision types captured by the AFC dataset also highlights that different regional and ideological models of academic freedom provisions have formed and spread over time. For instance, we find that the phrase “academic freedom” is not very commonly used in constitutions outside the hispanophone world and only really catches on in various African constitutions after Liberia’s precedent in 1986. Yet at the same time, a second model emerges in other African countries, starting with São Tomé and Principe in 1975. A closer examination of their provisions shows that they reference the freedom of “scientific creation”, “creativity” or similar—a phrase that can also be found in many constitutions in Eastern Europe and Central Asia before and after 1989. These provisions reflect the strong socialist influence on many constitutions of the twentieth century, and many constitutions even after the fall of the Soviet Union inherited these clauses.

The reasons why and circumstances under which this phrase was originally coined by Yugoslavia in 1963, later taken up by the USSR in the 1970s, and further spread to other world regions, remain yet to be explored in more depth. These and other patterns thus illustrate how the data provided in the new AFC dataset can also serve as a valuable starting point for more qualitative and in-depth research into the origins and circulation of academic freedom as a constitutional right. The same goes for more normative questions, such as whether the different provision types also connect to differences in how academic freedom is implemented around the world. The Latin American case speaks most clearly to this question, where the constitutional emphasis on university autonomy is indeed reflective of an approach to academic freedom that is significantly different from that of other regions as it emphasizes the institutional rights of universities over individual freedoms of academics (Bernasconi 2023). Whether systematic differences can also be found for countries referencing “academic freedom”, as opposed to the freedom of science or other phrases, remains to be investigated. In this way, the empirical mapping provided by the AFC dataset can inform more theoretically and normatively oriented debates about the content, meaning, and historic emergence of academic freedom as a legal and organizing principle in higher education research and teaching.

The analysis of constitutional texts has some clear limitations. While the dataset comprehensively covers national constitutions that were in force over more than 230 years, it can capture neither the de facto protection of academic freedom on the ground, nor other de jure protections at legislative or subnational levels. That said, the AFC dataset can be a beneficial element in such studies as it allows to identify interesting discrepancies between constitutional protections and de facto infringements (e.g. by comparing it to the Academic Freedom Index, see Spannagel and Kinzelbach 2022). It can also serve as a basis for studying how public and private higher education institutions are differently impacted by constitutional academic freedom guarantees, or to what extent the prevalence of private institutions may account for the absence of constitutional guarantees. Lastly, future data collection efforts may focus on complementing the AFC data with systematic analyses of academic freedom provisions in subnational constitutions in federal systems and in higher education legislation.

Conclusion

The AFC dataset provides the first systematic global resource that maps the historic development of academic freedom as a constitutional norm and its regional and international diffusion. The data show interesting patterns and trends in the adoption of constitutional academic freedom provisions, which lend themselves to various types of future analyses in political science, comparative constitutional law, and higher education studies. In particular, the functional connection of academic freedom provisions to the development of the higher education sector seems to be an important aspect that deserves attention. The separate consideration of different provision types further allows to retrace and examine distinct genealogies in the spread of academic freedom provisions. This includes, for instance, the important role played by socialist influences, which significantly contributed to putting the originally German provision of the freedom of science on the world’s constitutional map. Researchers can use the AFC data to do quantitative analyses, but they may also find them useful to systematically identify cases, such as the role of Yugoslavia, for more in-depth qualitative studies.