The Transparency of Constitutional Reasoning: A Text Mining Analysis of the Hungarian Constitutional Court’s Jurisprudence

The analysis of constitutional interpretation has received much attention in recent years. This article is a contribution to research using text mining methods to account for markers of constitutional reasoning in big data-sized text corpora. We examine how often the Hungarian Constitutional Court (the HCC) reflected on the various methods of interpretation. For this purpose, we have created a com - plex corpus covering all HCC decisions and orders between 1990 and 2021. We found evidence that the methodological practice of the HCC is not self-reflexive in general as only 44% of its decisions make a reference to at least one method of interpretation. We also show that the self-reflexive nature is even more prevalent (in fact, ubiquitous) in 100 doctrinally important decisions from the 30 years of jurisprudence in question. While this study is a first step towards the quantitative analysis of the reasoning of the constitutional judiciary, further mixed methods research is needed to account for intertemporal changes in such data and to refine the measurement of constitutional interpretation.


INTRODUCTION
In recent decades, the relevance of the judicial review has grown dramatically. 1It plays an increasingly important role in determining the direction, form, and content of constitutional law and policy in a growing number of countries.The methods of argumentation have become a pervasive feature of public discourse as well.The analysis of constitutional interpretation carried out in the first place by constitutional courts (and other high courts conducting judicial review) and the various methods used have received great attention in the scholarly literature. 2 Constitutional reasoning, understood as the justification given by constitutional judges of Law, 1-3 Egyetem tér, H-1053, Hungary; Rebeka Kiss, Junior Research Fellow, Institute for Political Science, Centre for Social Sciences, Budapest, 4 Tóth Kálmán u., H-1097, Hungary, and PhD student, University of Public Service (Budapest), 2 Ludovika tér, H-1083, Hungary; István Járay, Research Assistant, Institute for Political Science, Centre for Social Sciences, Budapest, It is generally considered that the HCC was "activist" in its practice, both in its powers and in its interpretation, during the period of László Sólyom's presidency.See B. Pokol, Constitutionalization  and Political Fighting through Litigation, "Jogelméleti Szemle" 2002, vol.1; S. Zifcak, Hungary's  Remarkable, Radical Constitutional Court, "Journal of Constitutional Law in Eastern and Central  Europe" 1996, vol. 3.This means that the HCC interpreted several abstract constitutional provisions as conferring jurisdiction on itself and went beyond the statutory rules in certain areas of its jurisdiction.Interpretative activism can be understood as a frequent departure from the constitutional text.The HCC was also criticized for creating new rules by an interpretation that was not present in the text or by not developing well-founded reasonings for one or other decisions.See A. Jakab, V.Z.Kazai, A Sólyom-bíróság hatása a magyar alkotmányjogi gondolkodásra, [in:] Kontextus által világosan: a Sólyom-bíróság antiformalista elemzése, eds.T. Győrfi, V.Z.Kazai, E. Orbán, Budapest 2022,  pp.115-137; K. Kovács, G.A. Tóth, Hungary's Constitutional Transformation, "European Constitutional Law Review" 2011, vol.7(2), pp.183-203; L. Sólyom, Introduction to the Decisions of the government -again gave rise to several debates on constitutional interpretation in the domestic literature. 6Taken together, the varying levels and focal points of the methodological self-reflection of the constitutional reasoning of the HCC serve as fertile ground for an analysis of interpretative practices over a longer time frame. 7n recent years, the practice of the Constitutional Court in Hungary and elsewhere has been examined qualitatively and quantitatively with different focuses and approaches. 8The precursors to our present analysis tend to agree that the HCC did not show a high level of explicit methodological self-reflection in its reasoning practice.In our understanding, derived from the available literature, self-reflection is the way the constitutional court reflects on its own interpretative activity, especially how it reflects on using one or the other method of interpretation when applying the text of the constitution. 9Doctrinally relevant landmark decisions (what we call the top 100 HCC decisions below) prove that it is part of the legal culture in Hungary to make explicit linguistic references to the applied method of interpretation. 10nstitutional Court of the Republic of Hungary, [in:] Constitutional Judiciary in a New Democracy: The Hungarian Constitutional Court, eds.L. Sólyom, G. Brunner, Michigan 2000. 6 A. van Aaken, C. List, Deliberation and Decision: Economics, Constitutional Theory and Deliberative Democracy, Ashgate 2004; L. Sólyom, The Rise and Decline of Constitutional Culture in Hungary, [in:] Constitutional Crisis in the European Constitutional Area, eds. A. von Bogdandy, P. Sonnevend, Oxford 2015.7 In this respect it does not matter if the decision is based on the Act XX of 1949 of the Constitution or on the Fundamental Law of Hungary after 2012.
10 See Az Alkotmánybírósági Gyakorlat I-II.Az Alkotmánybíróság 100 elvi jelentőségű határozata 1990-2020, eds.F. Gárdos-Orosz, K. Zakariás, Budapest 2021.In this article, we suggest that if the individual judge and therefore the court itself is self-reflective on its activity, we will see the linguistic signs in the decision itself, especially in the reasoning part.Regarding this definition, it is important to note that self-reflexivity and its linguistic presence are not a strict normative requirement for a legitimate decision; however, as part of the legal culture, it is very often present in the reasoning to explain the mindset of the judge(s).Our research questions in this article, therefore, concern the extent to which such observable self-reflecting methods of reasoning are present in the HCC's jurisprudence as well as the prevalence of individual methods and their dynamics.
In order to be able to conduct a quantitative analysis of a large sample of HCC decisions we collected a database which contains the decisions and orders of the HCC from 1990 to 2021.The database includes 5,336 decisions and 5,427 orders (the database includes all decided cases that reached a judge and were published in the official journal of the HCC -"Alkotmánybíróság Határozatai").We also analysed what was selected by experts to be the 100 most significant, "landmark" decisions of the HCC of this period. 11We investigated two hypotheses in relation to this database.Hypothesis 1 states that at least 51% of all HCC decisions carry an explicit reference to at least one method of interpretation.Hypothesis 2 posits that a sample of 100 landmark decisions carries more explicit references to at least one method of interpretation per decision than the count for the full sample of decisions.The first hypothesis is rooted in extant literature and an understanding that Hungarian legal culture puts an emphasis on proper judicial reasoning in jurisprudence.The second hypothesis is based on the assumption that the HCC goes out of its way to make sure this convention is upheld for what the legal community considers to be landmark decisions.
The research design applied made use of the counting of various versions of keywords that can be attributed to a number of reasoning methods which we derived from the literature.We also validated the matches extensively to make sure that only good matches were counted.
Results show that the majority of HCC decisions did not feature even a single explicit reference to one of the constitutional reasoning methods under consideration.The sample of 100 landmark decisions, however, shows a decidedly higher prevalence of such markers of constitutional reasoning.
In what follows we first present a review of the relevant literature.Next, we outline our theoretical framework, formulate the research questions and the hypotheses.
The following segment describes the dataset, and the quantitative empirical research methods applied.The section on results presents and interprets statistics related to the prevalence of markers of constitutional reasoning in the corpus at hand.The final section concludes and discusses avenues for future research.

LITERATURE REVIEW
In this literature review, we first consider the publications that provide a comparative analysis of the constitutional courts.Then we focus on works which examine the HCC from different perspectives.The discourse on constitutional law has its own specific language.As A. Jakab argues, the task of constitutional theory is to discover this language.12In the monograph European Constitutional Language, he outlines the foundations of constitutional interpretation and statutory interpretation in European and continental law, as well as the distinctive features of the style of reasoning of ordinary and constitutional courts, and reviews the various methods of interpretation. 13he principles of constitutional interpretation in continental law are very similar to classic legal interpretation,14 with certain specific features, as is well described in the work by F. Gárdos-Orosz and Z. Szente on the art of constitutional interpretation.15J.P. Dawson and G. Gorla have compared the brevity and rhetorical style of French Supreme Court opinions with the more discursive approach taken by American judges. 16The research by M. de S.-O.-l'E.Lasser and N. Huls, M. Adams, and J. Bomhoff also contributed to the analysis of the latter approach. 17By comparing the reasoning of judges in the French Cour de cassation, the US Supreme Court, and the European Court of Justice, Lasser's analysis seeks to cast a broader light on the wider discursive context in which these judges pronounce their decisions.He explicitly points out that what he says about the reasoning practice of the Cour de cassation applies equally to the Constitutional Council (Conseil constitutionnel).
More recent research, such as the comparative work of J. Goldsworthy, has begun to look specifically at how constitutional reasoning differs between constitutional systems by providing country studies. 18A methodological update to the research on constitutional reasoning is the study of A. Jakab, A. Dyevre, and G. Itzcovich in which the authors point out that studies focusing specifically on reasoning tend to be purely analytical or normative without comparative and/ or empirical perspectives on constitutional and related sociological issues. 19In another edited volume titled Comparative Constitutional Reasoning, the authors highlight the world's leading independently reviewed cases through a combination of qualitative and quantitative analyses (yet do not utilize text mining techniques to conduct their analyses).
Legal scholars in general, however, focus rather on how judges should arrive at their decisions in the light of what they actually do. 20The currently available national and international literature points to the fact that only a few studies on legal and constitutional reasoning apply quantitative methods.The analysis of legal texts by different methods has a long history, but for a long time, the field has been dominated by qualitative methods alone.The use of less traditional quantitative methods, such as text mining, has appeared in Hungary in the social sciences, similar to international trends.Since the 1990s, legal texts have increasingly been seen as data, and by using this method, previously unexplored phenomena can be made more understandable to researchers.
The text mining method is based on various data analysis algorithms to process unstructured textual data sets.Much of the information of interest to lawyers, jurists, and legal scientists is presented in the form of texts, whether they are pleadings, actions, contracts, court decisions, law journal articles, legislative acts, or constitutional court decisions.For centuries, the search for and analysis, comparison, and 18 See Interpreting Constitutions… 19 A. Jakab, A. Dyevre, G. Itzcovich, Conreason -the Comparative Constitutional Reasoning Project: Methodological Dilemmas and Project Design, "MTA Law Working Papers" 2015, no. 9, pp. 3-23. 20 For example, see V.C. Jackson, Multi-Valenced Constitutional Interpretation and Constitutional  Comparisons: An Essay in Honor of Mark Tushnet, "Quinnipiac Law Review" 2008, vol.26(599)  Jakab, D. Kochenov, Oxford 2017, pp.28-43; S.M. Griffin, American Constitutionalism: From  Theory to Politics, New Jersey 1996, pp.140-191.interpretation of these documents has been the task of legal practice and jurisprudence.Lawyers deal with words.
A. Dyevre explains that while the study of legal texts is as old as legal science, what is new is the emergence of a whole range of text-mining techniques for analysing and processing data, which help lawyers, researchers, and the legal community to navigate, understand, and analyse the ever-growing sea of legal and legally relevant documents.These techniques rely mainly on recent advances in machine learning and language processing technologies. 21In this respect, the research of T. Groppi and M. Ponthoreau, which comparatively studies the use of foreign precedents by constitutional judges, is also noteworthy. 22With this monograph, we now have data on the actual number of cases citing foreign case law in 16 countries.Z. Szente contributed to the analysis of the HCC. 23here quantitative methodologies appear, they are often based on manual research on official websites and expert selections, and when based on machine learning, they typically use network research techniques.E. Bodnár's research encompasses various empirical methods, case law analyses, and expert interviews.The purpose of her study was to find all cases where the HCC referred to foreign law, including references to specific regulations, case law, or general concerns.The author used the public online database available on the website of the HCC for her research and applied manual counting 24 as the website's database is unsuitable for more profound text mining research.
Following international trends, applying various text mining methods and techniques in social sciences and law has also gained ground in Hungary.One of the prominent pioneers in this field is L. Blutman, who examines the methodology of legal analysis.In his study, he seeks to answer the question of the rules that govern the formation, justification, or critique of legal statements, using language-centric and empirical methods to conduct a scientific analysis of legal texts. 25Blutman's work is pioneering since textual empiricism in the study of analytical legal doctrines is new in current mainstream legal research and its traditions in Hungary. 26n his language-centred textual analyses, he assumes that individual legal norms are created through language and that only language itself can create legal norms in 21 A. the human mind.In law, many questions are decided by the linguistic expression of particular ideas, which is essential in establishing the chain of causality and rationality.When studying the proportionality test, Blutman draws attention to the fact that unfortunately, the court's language is sometimes inadequate and undefined. 27his appears as a difficulty in achieving conclusive results with text analysis. 28s.Ződi's study uses network research methods to analyse and examine the interferences of the decisions of the HCC between 1990 and 2017.His research highlights that the mapped reference network follows the same pattern as almost all court reference networks around the world analysed by network research methods.The research demonstrated that network science could be an exciting complement to doctrinal jurisprudence in that network science, like other quantitative-based sciences, can reveal regularities. 29lthough from an investigative point of view, K. Pócza, G. Dobos, and A. Gyulai take a new approach to the examination of the decisions of the HCC, as they have developed a text mining-based methodology for systematically mapping the multifaceted reality of constitutional adjudication by measuring the strength of judicial decisions. 30Another precursor of our present study in terms of quantitative research methodology is the volume in which Pócza and co-authors using an innovative research methodology, quantifying the impact and effect of judicial decisions on legislation and legislators, and measuring the power of judicial decisions in six Central and Eastern European countries. 31he analysis of the language of constitutional law in the HCC decisions, together with the explicit textual analysis of the legal justification of HCC decisions, have so far mostly been conducted by applying expert-based research methods.Therefore, in this study, we use a new approach -text mining.In doing so, we depart from the empirical research methods used in the past on order to support them with new findings.

THEORETICAL CONSIDERATIONS AND HYPOTHESES
Studying constitutional reasoning has produced a massive literature.Based on this scholarship we understand constitutional reasoning (or argumentation which we use as a synonym of reasoning) as a special type of legal reasoning. 32We define interpretation as the determination of the content of the normative text.Hence, interpretation is part and parcel of constitutional reasoning.As Jakab put it, "what is traditionally called 'a method of interpretation', is in fact a type of argument used to interpret a text". 33There is a difference between legal interpretation and legal argumentation (legal reasoning).Interpretation involves exploring the meaning and rationale of a legal norm in a specific case.Conversely, argumentation aims to justify the application of the norm in a particular way and with a specific meaning.Interpretation is a rational activity, while argumentation is a rationalizing process. 34aw uses various methods to make its decisions, and the canon of interpretation is a common form of this. 35Law is first objectified in writing as a text and then processed further intellectually.Understanding or interpreting a law produces a different set of meanings; these can be fixed for a long time. 36For this reason, it is essential that in a democratic society, all moments of understanding and enforcing the law must be public, and therefore moments of interpretation also require some sort of publicity.
In our research, we rely on the linguistic characteristics of "justificatory reasons" 37 in the context of reasoning, i.e. we examine the methods of reasoning that the HCC is called upon to use in the reasoning of a decision.In this study, constitutional reasoning is examined in its narrowest sense, 38 i.e. we focus only on the text of the constitutional reasoning 39 of constitutional courts.
The methods of interpretation themselves are generally not fixed by law but are developed by judicial practice.Where there are constitutional or statutory rules on the methods of interpretation such as in Hungary in the Fundamental Law, this is not a taxonomic list.The Hungarian domestic legislation is unique in that the constitutional legislator provides (incompletely and not exclusively) methods of interpretation in among others Article R of the Fundamental Law.
After the democratic transition, Hungary followed the German model of constitutional jurisprudence, the methods of interpretation were developed by the HCC.In Hungary's Article R of the Fundamental Law, the constituent power, however, declared that "the provisions of the Fundamental Law shall be interpreted by their purposes, the National Avowal contained therein and the achievements of our historical constitution.The protection of the constitutional identity and Christian culture of Hungary shall be an obligation of every organ of the State".
In addition to the provision of the Fundamental Law, Act CLI of 2011 on the Constitutional Court contains an explicit provision in Article 63 (2) concerning the statement of reasons for decisions which states: "With the exception of rulings with a summary statement of reasons as specified in Subsection (3) of Section 56, 40 the Constitutional Court shall be obliged to give detailed reasoning for its decisions".Similarly to the German model and emphasising the importance of interpretation, it is the possibility of adding concurring and dissenting opinions to the majority decision with alternative interpretation, which the Act on the Constitutional Court states in Article 66 (2) and (3) as follows: "If a member of the Constitutional Court who opposed the decision of the Constitutional Court is outvoted, he or she shall have the right to attach his or her dissenting opinion, with a written reasoning, to the decision", and "A member of the Constitutional Court who agrees with the merits of the decision shall have the right to attach his or her reasons in a statement if they differ from those of the majority".
We identify different methods of interpretation in constitutional scholarship.These are very well-known reflected patterns of reasoning acknowledged as rational and legitimate, which are used to limit the scope of the interpretation of the abstract rule and avoid arbitrariness during the concretization of the rule to the specific constitutional controversy, or to the particular constitutional question. 41alapjog-érvényesítés gyakorlata, ed.É. Balogh, Budapest 2019, p. 374; K. Zakariás, A bírói döntések alkotmánybírósági felülvizsgálata terjedelmének dogmatikai keretei -A jogalkalmazás közvetlen és közvetett alapjogsértésének kontrollja a német és magyar gyakorlat tükrében, "Állam-és Jogtudomány" 2021, vol.4(5), p. 106. 40In case of the rejection of admission, the panel shall pass an order that contains a short reasoning specifying the ground for rejection.
41 J. Goldsworthy, Constitutional Interpretation, [in:] The Oxford Handbook of Comparative Constitutional Law, eds.M. Rosenfeld, A. Sajó, Oxford 2012.This doctrinal framework, which we usually call the methods of interpretation, is based on experience and is simultaneously embedded in normative philosophical requirements. 42Suppose the judges would like to refer to the context of the provision in the text to find out its meaning.In that case, the legal doctrine evaluates whether it is acceptable to determine the definition of a piece of an abstract text by analysing its context.In this specific case, legal scholarship has found that this judicial practice is rational and therefore qualifies as a legitimate method of interpretation, identified as the contextual method of interpretation.
Some centuries ago, the accepted methods of legal interpretation became crystallized, and constitutional law, 43 at least as understood in the continental legal systems, adopted these methods and adapted them to the constitutional reasoning that remained in these systems -including the Hungarian -a fundamentally legal task.These methods and the legal nature of constitutional adjudication in the normative sense were crystallized after World War II, and doctrinal expectations regarding constitutional adjudication were formulated.These doctrinal expectations were emphasised not only by legal scholars working with a rational-legal doctrine (dogmatics according to German usage) 44 but also by political institutions.
The classical methods of interpretation were not named explicitly in the text of the Constitution of the democratic regime change of 1989.However, they were identified before the entering into force of the Fundamental Law by the HCC and by legal scholarship from the early nineties as the classical methods of interpretation.This paper will not provide a comprehensive and detailed description of the specific classical methods of interpretation that the HCC may use because several studies have already been written on the complexities of constitutional reasoning and the listing of correct and incorrect methods of reasoning and interpretation. 45For the sake of 42 L.L. Fuller, K.I.Winston, The Forms and Limits of Adjudication, "Harvard Law Review"  1978, vol.92(2), p. 354. 43The development of legal interpretation methods and their application to constitutional reasoning is an ongoing process in various legal systems.Scholars and jurists have refined these methods, incorporating historical, textual, teleological, and systematic approaches.While scholars such as Friedrich Carl von Savigny in 1840 have made significant contributions to the theoretical foundations, developing interpretive methods for constitutional reasoning involves scholars from different periods and legal systems.The classical methods of interpretation, as established by Savigny, include textual, systematic, historical, and teleological interpretation.See F.C. von Savigny, Systems des heutigen Römischen Rechts I, Berlin 1840. 44Noteworthy is the German practice, where the Rules of Procedure of the Federal Constitutional Court require that, as a general rule, a written opinion (Votum) must be submitted in all Council (Senat) cases and only exceptionally, in simpler cases, a reasoned draft decision may be submitted ( § 23).This opinion is in effect a technical report, which summarises everything needed to decide the case and sets out the investigation step by step.This material can run to hundreds of pages for complex cases.See Federal Ministry of Justice, Rules of Procedure of the Federal Constitutional Court, http:// www.gesetze-im-internet.de/englisch_bverfggo/index.html(access: 30.8.2023).
45 F. Gárdos-Orosz, Z. Szente, The Art of Constitutional Interpretation… our study, measuring the self-reflexivity in the use of the methods of interpretation in constitutional court jurisprudence, we have chosen some most important methods and we aimed to prove with the above explanation that we can proceed with the text mining despite these and other differences between the two constitutional eras in Hungary before and after the entering into force of the new Fundamental Law in 2012.
In light of these considerations, below we list the methods that became important and recognizable in the jurisprudence of the HCC following the democratic transition.To take the period between 1990 and 2021 in one corpus for our examination, we will match the classical methods 46 with the new regulations listed above from the Fundamental Law of 2011.We delineate six such methods: linguistic, teleological, contextual, historical, "beyond the law", and decision-based methods of reasoning. 47irst, linguistic (or grammatical) interpretation is associated with pure textualism or the so-called direct meaning rule.It is not an explicit requirement of the constitutional text of the Fundamental Law, so we will search for the related words in both the 1989 Constitution and the 2012 Fundamental Law-based jurisprudence.
Second, the teleological (purposive) interpretation wishes to discover the goal of the provision.This emerges in Article R of the Fundamental Law, in which the requirement is that constitutional provisions should be interpreted in accordance with their purposes.We would think that this provision alone would give a wide margin of appreciation to the HCC, but the next sentence in Section 4 about the protection of constitutional identity and Christian culture restricts this freedom to defining the purpose of the rule.However, we could search the words and expressions related to the teleological interpretation both in the pre-2012 and the post-2012 jurisprudence.
Third, contextual interpretation occurs when the constitutional text is understood in the entire context of the constitution, considering the other related provisions of the text.The integrity of the constitutional text is a keyword in this method.Article R of the Fundamental Law requires the broad contextual interpretation explicitly, in the strict sense, and implicitly, in the general sense.In the strict sense, it requires that the Preamble called National Avowal of the Fundamental Law be considered when interpreting the other provisions of the text.This is a requirement of the coherent 46 F.C. von Savigny, op.cit. 47F. Gárdos-Orosz, Constitutional Interpretation under the New Fundamental Law of Hungary, [in:] Populist Challenges… An additional pragmatic interpretation occurs when the judge considers the decision's social, economic, technological, political, etc., effects.Article N of the Fundamental Law requires all state organs to act with respect to the financial goals of the state.However, this provision does not have significant relevance, as according to Article 37 (4), the Constitutional Court cannot review controversies related to public finance legislation.Still, if the Constitutional Court must observe the financial goals of the state, the necessity of the pragmatic approach to the constitutional interpretation becomes a requirement.Search words related to this interpretation are therefore valid both before and after 2012.interpretation of the constitutional text, which includes the preamble, i.e. the long National Avowal with the values of the political majority contained within it.
In the broad sense, we argue that when a contextual analysis is carried out on the Fundamental Law, it is not restricted to the constitution itself, according to Article R of the Fundamental Law historical constitution and Christian culture should be considered the context of the entire Fundamental Law.While we emphasise the theoretical importance of this provision, in the practice of the HCC in the examined period, this latter contextual understanding has not yet gained relevance and therefore we could use the same set of search words for both periods.
Fourth, Article R of the Hungarian Fundamental Law refers to the achievements of the historical constitution as a reference point for interpretation.It emphasises the long history of constitutional values in Hungary.It operates in the constitutional jurisprudence after 2012 by mentioning the achievements of the historical constitution that could be otherwise understood as a pure historical method of interpretation. 48Therefore, we connected the reference to the achievements of the historical constitution with the other search words related to the historical interpretation in the former and in the present constitutional jurisprudence.
Fifth, the beyond the law or moral interpretation is based on the assumption of a political philosophy behind the constitutional text, leading the judge to a morally correct understanding of the norm.This political philosophy is based on the community's morals in constitutional populism.The necessity of the moral sense is also present in the text of the Fundamental Law, when, e.g. in Article R, the Fundamental Law requires respect for constitutional identity.
As the notion of respect for constitutional identity was not previously defined in the constitutional text or elsewhere, it did not have a legal meaning at the moment of adoption (although it did have a political one); therefore, there is -in a theoretical sense -a textual window to allow the political philosophy of the constitution-making majority to become one of the tools of interpretation.Prior to 2012, the moral interpretation was rather based on the Kantian understanding of morality that guides the decisions of constitutionality.No matter how the content is different before and after 2012, the moral interpretation as such is a legitimate, acknowledged way of classic interpretation, therefore it can be examined on the entire corpus.
Finally, we define decision-based interpretation methods as those referencing former decisions.Hungarian Constitutional Court decisions do not constitute precedent in the classical (common law) sense.However, Z. Pozsár-Szentmiklósy points out that the HCC cites its own relevant practice -in most cases citing the findings of the "reference case".This method aims to highlight the coherent practice of HCC.The literature points out that the advantage of this interpretation method is that it enhances the transparency of the structure of the reasoning of decisions and their persuasive power.This is necessary to increase public confidence in the HCC's activities. 49n our research design, we consider an HCC decision as self-reflexive if it includes at least one reference to any of the methods of interpretation in its reasoning.In order to understand the usage of constitutional reasoning in the jurisprudence of HCC we examine the following hypotheses: H1: At least 51% of all HCC decisions carry an explicit reference to at least one method of interpretation.
H2: The sample of 100 landmark decisions carries more explicit references to at least one method of interpretation per decision than the count for the full sample of decisions.
The first hypothesis is rooted in extant literature and an understanding that Hungarian legal culture puts an emphasis on proper judicial reasoning in jurisprudence.The second hypothesis assumes that the HCC goes out of its way to make sure this convention is upheld for what the legal community considers to be landmark decisions.

DATA AND METHODS
We procured our data on the HCC's decisions from the website of the HCC, where officially published decisions are openly available. 50To verify our data -as all industry-standard legal databases obtain their data from the official HCC website -we also cross-checked it using different legal databases.The database contains all of the decisions and orders of the HCC from 1990 to 2021: 5,336 decisions and 5,427 orders (taken together: 10,763 decisions).Apart from the corpus of decisions, our database contains metadata related to each decision.
The decision texts from the initial database were pre-processed by removing all non-alpha characters (e.g., punctuation marks, numerals, roman numerals, etc.) and lowercasing them.Our analysis used two variables from the available metadata: the year of the decision and the list of citations of external legal documents.The year variable is an integer; the variables containing the cleaned texts, and citations are strings.We show the first two rows of our input data in Table 1.'1989.évi xxxi.törvény 16. paragrafus', '1989.évi  xvii.törvény 10. paragrafus 2. bekezdés a. pont',  '1989.évi xvii.törvény 4. paragrafus', '1989.évi  xxxv.törvény', '1989.évi xvii.törvény', '1989.évi  xvii.törvény  We applied a mixed methods approach to examine this corpus relying on both qualitative and quantitative methods.First, we selected the keywords related to the specific methods of interpretation based on the above academic research on constitutional and reasoning.Second, we used text mining to measure the prevalence of keywords in the underlying corpus of HCC decisions.
Our methodology relies on a dictionary-based approach.We counted the instances of keywords in every document and aggregated the number of keywords matches in the text (see Appendix A for a slew of examples related to various logics of reasoning).We examined the keyword matches in every category of methodological self-reflection.We also did a statistical analysis of the total of keyword matches.Finally, we normalized the number of counts by the decisions' token lengths (i.e., word counts).In our analysis, we refer to this normalized count index as the Count Index.
The identification of search terms was conducted on several levels.The selection of words was based on the literature,51 so that the decisions and the keywords highlighted in previous research analysing the practice of the HCC are included in our dictionary.In addition, we have carefully investigated the expert selected 100 important decisions and highlighted the words used in them that describe methods of interpretation. 52Table 2 presents the six categories of reasoning and the associated keywords.Alkotmánybíróság eddigi gyakorlatában * However, with regard to the external validity of our findings, it should be emphasised that dictionaries need to be tailored to other jurisdictions and contexts, and validated appropriately in that context, in a similar way to the methodology used in this article.
** The Hungarian version refers to the explanatory note of the "law", as this is what is used in practice (e.g., in the text of the HCC decisions).Yet in reality, the explanatory note is associated with bills.
Note: we have made a distinction between the terms "Constitution" and "Fundamental Law" due to the entry into force of the Fundamental Law in 2012.
One marker of constitutional reasoning was referencing Article R (3) of the Fundamental Law.In this specific case to be as accurate as possible, we first filtered for those observations containing a reference to this paragraph based on the lists of citations.Then, we summed the instances of phrases matching this paragraph in the text.To avoid inflating our results by unwanted matches, we took two steps.First, we searched for colloquialisms, excluding keyword matches where a part of the searched expression could be part of an expression with a different meaning.Second, we excluded a stop list of words and colloquialisms from the corpus before applying the counting.
The stop list initially consisted of the unwanted phrases related to each methodology listed in Table 1.We then augmented the stop list by manually filtering unwanted terms in those documents where the sum of matching keywords exceeded six.The list of augments consists of: igazságok, igazságügy, igazságszolg, igazságtart, történeti hivatal.(Due to the difficulty of accurate translations, we included the original Hungarian list.) The list contained in Table 2 is certainly not a closed canon of methods of interpretation and even less so a definitive list of associated words and expressions.Still, the above compilation (and additional rules) is rooted in a qualitative analysis of actual decisions, Fundamental Law requirements and a reflection in practice that has emerged in the literature, classifying and identifying the various methods of interpretation that the HCC can use to reach its decision.A manual validation of each and every individual automatic match was also applied with non-relevant (the filtered and excluded words based on the stop list) matches excluded from the final tally.
An additional methodological remark is that we make no claim that if the HCC does not name one or the other method by the words identified above it is not engaged in constitutional reasoning.We argue, however, that by using this text mining method, we could discover approximately how often and in which cases the court was explicitly self-reflective of its use of one or the other method by using those words and expressions that are widely known in legal scholarship.

DESCRIPTIVE STATISTICS
The following section includes a comparison of the select top 100 HCC decisions with the remaining 10,663 decisions out of the total of 10,763.The token length of individual decisions (i.e. the word count of pre-processed texts) ranges from 43 to 22,054; the standard deviation is slightly above 2,000, and the mean is above 1,700.Table 3 shows the descriptive statistics of the document lengths and the total counts.The document length distribution is uneven, with little more than 2,000 tokens at the 75 th percentile and a maximum of above 22,000.The total number of keyword mentions is generally low, with the 75 th percentile containing one match (most decisions had no more than one keyword match).We can observe a significant disparity when comparing the document lengths among the landmark 100 HCC decisions and the rest of the corpus.Table 4 compares the descriptive statistics of the two sub-groups, and Figure 1 displays the distributions of document lengths.Generally, the distribution of the top 100 HCC decisions is more skewed to the right than the rest of the corpus, ranging from 1,039 to 22,013, instead of the minimum token length of 43 and maximum of 22,054 among the other decisions: landmark decisions are longer.They also contain significantly more keywords on average.Table 5 shows a category-by-category description of our corpus.A significant difference is observable regarding the proportions of documents with at least one keyword of any category between the two groups: 99% of the expert selected, doctrinally most important decisions contained keywords, whereas this proportion is only 44% for the whole corpus of HCC jurisprudence between 1990 and 2021.As to the referred methodology, our results show that decision-based on former decisions methodology is the most prevalent in the entire corpus, with 30% of documents containing related keywords, followed by references to contextual argumentation (21%). 53Also, among the top 100 decisions, references to contextual methodology are the most frequent (75%), followed by decisions based on former decisions (64%), then teleological (48%) and historic (48%) references are made the most frequently.Across all categories, the proportion of documents with at least one keyword is significantly higher among the top 100 than in the entire corpus.Figures 2 and 3 elaborate on the faceted yearly distributions of the means of count indexes, supporting the cross-sectional difference in references to constitutional reasoning shown in Table 5.

Methodology: Linguistic
Methodology: Teleological Methodology: Contextual 53 The high proportion of contextual interpretation is not unexpected, as it is a method of interpretation that is often used in conjunction with other methods of interpretation.J.Z. Tóth analysed the legal interpretation activities of the European Court of Human Rights.In his general findings, he pointed out that contextual interpretation, in the broadest sense, is a method used in the vast majority of cases and can be found in the vast majority (almost all) of the decisions he scrutinised.See J.Z. Tóth, Interpretation of Fundamental Rights…, p. 78.In sum, less than 50% of HCC decisions contain self-reflective keywords (44%) in the complete corpus of the HCC jurisprudence.In contrast, 99% of the top 100 decisions have at least one mention of a searched term, indicative of self-reflective reasoning (see details in Table 5).The comparison of methodological references showed that mentioning the keywords decision-based on former decisions was prevalent among both groups.Still, the use of the keywords contextual, teleological, and historic were decidedly more likely to be used in the top 100 most important landmark decisions.
Looking at the rank order of decisions with the most keywords, we see a disparity in the results, whether sorted by the total number of keywords matches or the Count Index.Tables 6 and 7 show the two top lists up to 15.We sorted Table 6 by total counts and did not apply restrictions.The top list contained ten decisions which were part of the top 100.In Table 7, we only included those documents which have a length of over 3,000 words, sorted by the Count Index.We did so to decrease the bias towards shorter decisions in the top list where the denominator of document length is small.Only 4 of the 15 observations were part of the list of top 100 decisions.The disparity is not surprising if we consider that the distribution of document length in the top 100 is much higher than in the entire corpus.

RESULTS
Besides exploring our dataset in terms of its descriptive features, we also investigated two hypotheses related to the theoretical literature.Our first hypothesis set the reference threshold for the practice of the HCC to be generally considered self-reflexive as 51% of decisions containing at least one keyword.Our second hypothesis expected the sample of top 100 decisions to contain more markers related to constitutional reasoning as the rest of the corpus.Figure 4 shows a side-by-side comparison of the distribution of the proportion of documents containing keywords in a given year.On the one hand, among the top 100 HCC decisions, every decision has at least one self-reflective keyword in most years (the sole exception is 2007).On the other hand, the same proportion is mostly below 50% among the remaining HCC decisions, despite an increasing trend since 1990 and surpassing 50% from 2015.All in all, the empirical analysis of explicit references to types of constitutional reasoning lends support to both of our hypotheses.An important caveat is the dynamics of these averages which reveal important temporal differences between various periods.
To control for the impact of the length of individual decisions we also compared the normalized distributions of the Count Indexes between the two groups (see Figure 5).A dominant spike shows the overwhelming lack of self-reflective keywords in the HCC's argumentation in the entire corpus.However, the distribution of the top 100 decisions appears to be much flatter and skewed to the right, signalling more decisions containing more keywords regardless of length.
This cursory analysis offers a first quantitative glance on the observable markers of constitutional reasoning in the jurisprudence of the HCC over more than three decades.Further research on other country cases could reveal whether the Hungarian case is the exception or the norm in terms of the practice of explicit constitutional interpretation in decision texts.Similarly, within case comparisons (such as more systematic analysis of different periods in the composition and leadership of the court) could shed light on not just the trends in constitutional reasoning in general, but also on the limits of the research design proposed in this article.

CONCLUSIONS
Constitutional reasoning is a critical aspect of the jurisprudence of constitutional courts.Yet despite its importance, in extant research only a few studies apply generalizable, quantitative frameworks to the study of this aspect of legal reasoning.Where quantitative methodologies are utilized, they are based on manual data collection on -mostly -subsamples of the full body of decisions based on time limitations and/or expert sampling of "important" decision.
In the article, we argued that constitutional courts use methods of interpretation to explain their decisions.It is often assumed and required that courts have a conscious and self-reflecting, visible and, therefore, transparent reasoning practice in a normative sense.As the decision process has a linguistic manifestation, the constitutional court gives a public account on its the reasoning.This often includes the reference to the methods of interpretation that were applied by the court.
In this study, we investigated which methods of constitutional reasoning and how often were referenced in the jurisprudence of the HCC between starting from the regime change 1990 (and thus covering the democratic period where judges had real autonomy in making and discussing their decisions).In our quest to answer this research question we offered a three-fold contribution to the literature.First, we crafted a quantitative research design (rooted in a rigorous review of qualitative works) for a subject mostly analysed with a doctrinal logic that is replicable and scalable to other context for the comparative study of constitutional reasoning.Second, we applied that methodology to a new dataset of over 10,000 decisions of the HCC spanning more than three decades.Third, we offered a first, mostly descriptive statistics-based examination of the prevalence of explicit linguistic markers related to various forms of methods of interpretation in the dataset at hand.
Our results show that practice of the HCC is not overwhelmingly self-reflexive with 44% of decisions containing at least a single reference to keywords associated with logics of constitutional interpretation.In so far as the composition is concerned, we found that the HCC often based its decisions on values and interests beyond the constitutional text by using non-legal interpretation methodology.We also examined and compared these results with the references in the expert selected top 100 landmark decisions of the HCC.Here, we found that these decisions of the HCC are more self-reflexive than the rest of the sample with almost all key decisions containing references to at least one method of interpretation.Thus, we established that the HCC makes a more concerted effort to provide explicit arguments for decisions of legal doctrinal significance.
Our methodological approach is certainly not without its limitations and can only be considered to be a first step towards a fully-fledged, mixed methods approach to account for trends in constitutional reasoning for individual courts.Here we raise two such limitations which should be explored further in future studies.A general limitation is related to the role of explicit linguistic markers in constitutional reasoning.One might claim that constitutional interpretation is subtler and can only be deciphered by "reading between the lines".We partly confronted this argument by doing a qualitative analysis of scores of decisions as we looked for suitable keywords.We were convinced that such keywords exist but also readily accept that context matters beyond words and phrases and further efforts should be directed at understanding the linguistic representation of such reasoning better.
Secondly, and more case-oriented, our empirical results revealed major differences between individual periods in terms of the prevalence of keywords.This may reveal structural forces (such as the role of court composition) at play that should be investigated further.A corollary to this point relates to the disambiguation of token references to such reasoning and substantively applied ones.A legitimate case can be made that the higher keyword values for the post-2010 period do not betray a higher level of self-reflexion given the overall illiberal nature of the Orbán regime (and the effect of its court packing activities).Such issues can only be negotiated on a case-by-case basis and warrant the splitting of longer time frames in any research design (as we did in this article).
Despite these limitations, we do not see any major obstacles to measuring and comparing the practice of other constitutional courts of states with similar legal systems based on the methodology presented in our study by tailoring the dictionary to other legal systems and contexts, and validating it appropriately in that context.Such a future comparative analysis could provide deeper insights into the differences in the culture of judicial reasoning in different legal systems.With additional studies, we can establish historically and legally relevant benchmarks for the level of constitutional self-reflection and properly situate the results presented in this study within general trends of constitutional jurisprudence.parties, on the unconstitutionality of Act CXIX of 2010 amending Act XX of 1949 on the Constitution of the Republic of Hungary.Typically, the petitioners required its annulment ex tunc, i.e. with retroactive effect to its promulgation, because it introduced retroactive taxation in the public sector that was against the rule of law and legal certainty.
Although this was an important decision in terms of content, 54 here we are not concerned with the examination of the content, but with the methods used by the HCC in its reasoning.In the 61/2011.(VII.13.)HCC decision the linguistic, contextual, teleological, historical, "beyond the law" and "according to precedent interpretation" aspects are also reflected, along with the following keywords: linguistic, linguistic interpretation, the text of the Constitution, tradition, values of the Constitution, legislative purpose, beyond the law, extra-legal, constant practice of the HCC. 55In Table A1 we present examples of the interpretative methods used in the 61/2011.(VII.13.)HCC decision.latban, "Jura" 2012, vol. 1, pp. 37-44. 55 In Hungarian the following keywords appear in the reasoning of the decision: nyelvtani, nyelvtani értelmezés, Alkotmány szövege, tradíció, Alkotmány értékrendje, jogalkotó célja, jogon túli, jogon kívüli, Alkotmánybíróság állandó gyakorlata.
49 Z.Pozsár Szentmiklósy, Precedents and Case-Based Reasoning in the Case Law of the Hungarian Constitutional Court, [in:] Constitutional Law and Precedent International Perspectives on Case-Based Reasoning, ed.M. Florczak-Wątor, London 2022, pp.116-117. 50For a detailed description of the database, see M. Sebők, R. Kiss, I. Járay, Introducing HUN-COURT: A New Open Legal Database Covering the Decisions of the Hungarian Constitutional Court for Between 1990 and 2021, "Journal of the Knowledge Economy" 2023.

Figure 1 .
Figure 1.Comparison of the length of landmark decisions and the rest of the corpus Source: Authors' own elaboration.

Figure 4 .
Figure 4. Proportion of decisions with keywords by year Source: Authors' own elaboration.

Figure 5 .
Figure 5.Comparison of the normalized distribution of the count index Source: Authors' own elaboration.
, pp. 599-670; G. Itzcovich, On the Legal Enforcement of Values: The Importance of the Institutional Context, [in:] The Enforcement of EU Law and Values: Ensuring Member States Compliance, eds. A.

Table 1 .
The first two rows of the input table

Table 2 .
The methods of reasoning and the associated keywords *

Table 3 .
Descriptive statistics of the corpus Source: Authors' own elaboration.

Table 4 .
Comparison of descriptive statistics in the two samples

Table 5 .
Comparison of keywords by category Source: Authors' own elaboration.

Table 6 .
Top list of decisions based on the count of keywords Source: Authors' own elaboration.

Table 7 .
Top list of decisions based on the normalized count index Source: Authors' own elaboration.

Table A1 .
Methods of interpretation found in the 61/2011.(VII.13.)HCC decision and examples of such methods 54T.Drinóczi, Gondolatok az Alkotmánybíróság 61/2011.(VII.12.)AB határozatával kapcso- Beyond the law Beyond the lawThis basic legal value ensures first and foremost that the law gives the measure and form to the exercise of state power, and therefore, it is not possible to give an arbitrary and beyond the law interpretation to the state interest