Supporting Comparative Studies of Judicial Behavior: Introducing the Australian High Court Database

Abstract

Comparative research on law and legal institutions depends on high-quality data infrastructure. This article introduces the Australian High Court Database—a new resource that encodes structured information on all full judgments of the High Court of Australia between 1995 and 2020, and all leave applications (Australia’s equivalent to petitions for certiorari) from 2003 to 2018. The database is built in accordance with core principles that support comparative research: it is adaptable, and comparable. By attending to jurisdictional specificity while adhering to general standards, the database supports both within-country analysis and cross-national comparison. We illustrate how the Australian High Court Database can be used to study comparative judicial behavior by analyzing judicial dissent rates across apex courts, judicial ideology, and agenda setting.

Introduction

Comparative research on judicial behavior has developed rapidly in recent years (Epstein et al., 2024) with studies proliferating beyond the US Supreme Court Database (SCDB) (Spaeth et al., 2024). Underpinning these new empirical studies is the development and collection of new databases on judicial behavior, which have now been made available in several jurisdictions, including Germany (Engst et al., 2020; Hamann, 2019), Israel (Weinshall & Epstein, 2020), and the United Kingdom (Hanretty, 2020a), while several other studies provide replication datasets (Ash et al., 2025; Nie et al., 2022; Skiple et al., 2021). However, until now, empirical legal research on the High Court of Australia has been stymied by lack of a systematic, reliable, and publicly accessible database. One existing source of public data on the High Court of Australia is the National Science Foundation High Courts database (Haynie et al., 2007). This existing database has been used in several studies of the High Court (e.g., Gill, 2002, 2009; Gill et al., 2011; Haynie et al., 2013; Sheehan et al., 2012) but was designed as a cross-national source of data and so contains less nuance than is available in a purpose-built in-country database (Lynch, 2014). The purpose of this article is to complement this emergent literature by presenting the Australian High Court Database (HCDB). It contains structured information on all full judgments of the High Court of Australia between 1995 and 2020, and all special leave applications—the process of judicial agenda setting, elsewhere referred to as petitions for certiorari—from 2003 to 2018, and will continue to be updated. By going beyond the limitations of existing sources, we address the concerns that have undermined the credibility of High Court research in the eyes of legal scholars in Australia and elsewhere.

The database was created to encourage and support cross-national comparison in research on judicial behavior. Here, we describe the development of our database on the High Court of Australia, its design characteristics, its theoretical underpinnings (supporting the strategic model of judicial behavior), and its application as a resource for the study of judicial behavior in Australia, domestically, and comparatively.

Best data design practices embrace several principles for improved development of data infrastructure on judicial behavior: (a) the ability to address real-world problems (defined by contribution to public discourse on Court behavior); (b) accessibility (free access online, well-developed documentation, and logical citation principles); (c) reproducibility (making not only the data available, but also the means to reproduce and extend the data, whether via instructions to human coders, or via automated content scraping); (d) sustainability (avoiding data exuberance, and exploring automation); and (e) extensibility (the data should support the extension of ‘add-on’ datasets for further research). ¹

Our database, which adheres to these principles, is available online at aushighcourtdatabase.org and the Australian Data Archive in several formats, including Excel, an R package, Stata, SPSS, and via an online analysis tool. Our extensive documentation allows reproduction: it includes detailed instructions to present and future coders on how to interpret the meaning of a variable, its source if an analogue is available elsewhere, and any relevant adaptations or clarifications. Our documentation supports sustainability by defining ‘core’ and ‘extended’ releases, to allow flexible data collection for future releases. Finally, our data are extensible, as our linked data on legal representation and special leave to appeal applications (petitions for certiorari) demonstrates.

Scholars of comparative institutions are aware that comparative data projects imply trade-offs (Freiburghaus, 2024). On the one hand, researchers should be mindful that blindly copying over coding principles to the next project can lead to pitfalls and misinterpretation. Indeed, as Weinshall and Epstein (2020, p. 418) argue, local jurisdictional expertise is required before a ‘single datum’ is collected. Additionally, data collection reflects theory, which in the case of judicial behavior has largely been developed with reference to the United States legal system. However, excessive focus on ‘ground-up’ approaches to empirical research and data collection can hinder comparison, especially when prevailing theoretical frameworks are ignored. Of course, institutions, legal traditions, and the resulting behavior differ. Our task, then, should be to record what is comparable and explain what is not.

Until now, publicly available data on Australian courts have variously given too much weight to comparability, arguably compromising the value of the data as a resource for Australian practitioners (Lynch, 2014), or to granularity, such that data summaries such as those presented here are impossible. We have in mind the Australasian Legal Information Institute whose data are comprehensive and of enormous utility to legal practitioners, but whose holdings function rather as a library of catalogued output from courts. This database remains a key source of raw primary sources for our own database.

In what follows, we begin with a brief overview of the High Court of Australia. We then outline the development of the HCDB and its origin as an adaptation of the SCDB (Spaeth et al., 2024), the principles underlying our adaptation to the High Court of Australia and Australian legal system, and its extension through linked datasets to study related phenomena, such as agenda-setting. We demonstrate the utility of the data as a resource to study comparative judicial behavior, with analysis of judicial dissent rates across apex courts, judicial ideology, and agenda setting.

Brief Background to the High Court of Australia

The High Court is a constitutionally mandated court, which was established by the Australian Constitution at the time of Federation in 1901. Initially, the High Court functioned as an intermediate court of appeal, formally sitting above lower federal courts and the state supreme courts, but below the Judicial Committee of the Privy Council, which sat in London, and was Australia’s apex court. Later, the High Court was formally entrenched as Australia’s apex court; since the Australia Acts 1986 (Cth and U.K.), the High Court has been the final court of appeal for Australia on all questions of federal and state law (including constitutional, statutory, and common law matters).

The High Court comprises seven justices, including the Chief Justice. High Court justices in Australia are formally appointed by the Governor-General, acting on the advice of the federal executive. In practice, the Attorney-General (the government’s chief legal officer) plays a central role. While not legally required to do so, typically they consult widely with state attorneys-general, senior judges, and legal experts before recommending a candidate to the Prime Minister, who advises the Governor-General. At different points in the Court’s history, the Prime Minister has taken the dominant role in selecting the justices, particularly when the Prime Minister has a legal background (Leslie et al., 2021). Although most judicial appointments to the High Court have not been as overtly political as in the United States, there have been multiple instances in which appointments have been viewed as primarily politically motivated in the sense of being designed to alter the ideological balance on the Court (Evans, 2001). For instance, immediately before appointing Justice Ian Callinan, the conservative Liberal government declared they were looking for a “Capital ‘C’ conservative.”

The Australian Constitution originally provided for life tenure for High Court justices; however in 1977, following an amendment to the Constitution, all High Court justices have been required to retire at 70 years of age (see Blackham, 2016). There are instances in which justices have strategically timed their retirement, or sought assurance from the sitting Prime minister that the government in power would appoint a replacement who had their political views or background. For example, in the 1940s, Justices Rich and Starke are widely regarded as strategically remaining on the bench, despite being aged 87 and 70 respectively, until the election of the conservative Menzies government, because they did not want the Chifley Labor government to be in a position to replace them (Coper, 1987). In early 1975, Justice McTiernan, who was a Scullin Labor appointee in 1930 and a staunch Catholic, sought assurances from Labor Prime Minister Gough Whitlam that if he were to retire that Whitlam would replace him with a Catholic (Whitlam, 2001). However, the Whitlam government was dismissed before a suitable replacement could be found. McTiernan retired the following year and his successor (Justice Aickin) was selected by the conservative Fraser government.

Since 1984, the High Court has had control over its appellate docket by virtue of the introduction of “special leave,” which gives the justices power to grant or deny leave to appeal to the Court. Pursuant to the Rules of Court, which are made by the Justices of the Court, the High Court will typically only grant special leave where the case presents a question of public importance, a conflict within or between federal and/or state courts, or it is otherwise in the interests of the administration of justice. Applications for special leave are initially examined by a panel of up to three justices. Most special leave hearings however, are decided by panels of two. Prior to 2016, applications for special leave were predominantly determined after a hearing (i.e., oral argument) before the special leave panel; however since 2016 the panel has discretion to decide special leave based on hearing or papers (i.e., without a hearing). While most cases in Australia are subject to special leave, Section 75 of the Australian Constitution provides for certain cases in the Court’s original jurisdiction. These are cases which the Court must decide as judicial review on Commonwealth matters, though remittal to lower courts are allowed. Still, relatively few cases appear under original jurisdiction, and successive governments have made attempts to curtail litigation in this area (Noonan, 2013). ²

If special leave is granted, oral argument is scheduled to be heard by either the full bench of seven justices or a smaller panel of at least three justices. In practice, the full bench is reserved for the most important cases, particularly constitutional cases, while most appeals are heard by a panel of five justices (Popple, 2001). The Chief Justice decides the size and composition of the panel.

The Data at a Glance

Here, we give a summary of the datasets available in the present release of the High Court Database. A summary of the linked datasets is shown in Table 1.

Table 1.

Summary of the Australian High Court Database aushighcourtdatabase.org

Dataset	Description	Timespan	N. Variables	N. Observations
justice_bio	Justice biographical information for 23 justices	1981–2020	155	23
justice_decision	Case context and justice decisions, for 1,816 High Court cases at the justice-case level.	1995–2020	319	12,712
advocate_case	Comprehensive legal representation information for 1,755 High Court cases	1995–2019	15	11,423
special_leave	Context and outcomes for 8,412 applications to appeal.	2003–2018	48	8,412

Justice biographical data (justice_bio) contains background information for 23 justices appointed to the High Court of Australia, from 1981–2020. The data are split into five categories of variables. First, there are identification variables, including information on registries, the nominating prime minister, the outgoing justice, the date of nomination, whether or not the justice later became Chief Justice, and their pre-appointment ideology score (see Robinson et al. (2021)). Second are variables on personal background, including date and place of birth, gender, race, religion, details of retirement and upbringing. Third, we incorporate family information, including class and parental occupation, and number of children. Fourth, there are details on educational background from school to terminal university degree. Finally, we record employment history up to appointment to the High Court, including length and seniority of judicial service, years of experience at the bar, and political affiliations.

Justice centered decision data (justice_decision) contains the decisions of justices for each case in the High Court from 1995–2020 at the justice-case level. The variables are split into the following categories. First, there are identification variables, including our own HCDB identifier and a number of other citation formats, as well as a set of indicators for multiple matters (where one cited case in fact covers several distinct decisions, coded separately in our data). The second category covers background information, including the types of litigants involved, the origin of the case, and information on lower court decisions. Third, the data identifies the procedure through which the case came to the court, including the special leave process. Fourth, the data includes chronological variables, including the date of the decision, as well as the dates of oral arguments, and relevant contextual information. Fifth, there are substantive variables recording the issues raised by each case, the authority for each decision, and decisions on costs and other factors. Sixth, there are the decisions of the individual justices, coalitions, and panel characteristics, as well as ancillary biographical information for each justice.

Advocates data (advocate_case) contains information on advocates appearing before cases at the High Court 1995–2019. This information includes names, gender, race, seniority (as a solicitor, barrister, King’s/Queen’s Counsel or Senior Counsel), and government affiliations as solicitors general for states, territories or the federal government, ³ and their role (lead, arguing, non-speaking appearance, or written submission only).

Special leave data (special_leave) codifies information on applications for special leave to appeal. This dataset contains information on case identification, area of law, case context and outcomes in the lower court, context of the special leave application, the panel selected to review the application, the outcome, and links to full High Court cases where the application to appeal was successful.

Development

Best practice design for databases on Courts has been articulated in Weinshall and Epstein (2020) and more recently in Engst and Gschwend (2024). We aim to replicate most of the conventions outlined in these works, while maintaining a focus on ease of use, familiarity, and extensibility. As outlined in the introduction, Weinshall and Epstein (2020) focus on relevance, accessibility, reproducibility, sustainability and extensibility. When describing the features of the German Constitutional Court Database, Engst and Gschwend (2024) also add three other desirable qualities of a well-crafted database. First, that it is a normalized relational database, which reduces redundancy and is less prone to errors. Second, it is parsimonious, in the sense that it contains a small and sustainable number of variables. Third, it is flexible, available as a series of rectangular datasets rather than a single monolithic dataset and contains a well-developed and conceptually clear codebook.

We concur with most of these desiderata but differ on some. For instance, we do not provide a relational database as the primary mode of distribution, though we still present a series of linked tables that make this goal entirely feasible should other researchers want to proceed with normalization for their own purposes. This still allows for flexibility and customization using widely distributed software packages such as R, Python, or indeed with database query languages such as SQL and its variants.

Second, we have erred on the side of data exuberance rather than keeping our codebook short and limited, which we view as a shortcoming only in the sense that it implies additional costs to update. Still, the rich and varied data that we have gathered are useful to us, and may be useful to others, including as a sample or ‘test-bed’, should future researchers wish to explore and expand data collection beyond what we collect here. We acknowledge that data exuberance brings trade-offs and for this reason we also provide a smaller ‘core’ database codebook, which should make future maintenance of the database more feasible and sustainable. In the supplementary information, both the full codebook and core codebook for the justice-case centered data are included. In the core codebook, priority has been placed on “high endorsement” variables (those variables which are relevant across most cases), “high variance” variables (in which values are not excessively dominated by a single outcome), and “automatable” variables (those variables which may be readily automated or scraped in future data collection efforts).

The starting point for development of the Australian High Court database was the SCDB, continuously updated and expanded since 1984 (Spaeth et al., 2024). The importance of this database for the comparative study of judicial behavior is difficult to overstate—all subsequent attempts at data collection in this field are in some way influenced by it. Accordingly, the simplest way to achieve comparative compatibility across judicial behavior datasets is to adapt the database design principles followed by the SCDB compilers. In the Australian case, this applies to the structure of rectangular datasets (for example, the justice-case centered format is retained) and the flexibility of available file formats. ⁴

This approach improves comparability for the study of the US Court system and other courts and it also allows us—as far as possible—to reduce friction for the user, with data that is both flexible and familiar to scholars around the world. Not only are structural conventions and formatting retained, so too are the naming and content of many variables (for example, the coding distinction between disposition and disposition direction; or the issue, issue sub area, and issue area hierarchical classification of case topic).

Because we retain many of the conventions from the SCDB, we can also be extremely clear about the areas in which adaptations of concepts and measures are necessary to fully capture the reality of judicial behavior in Australia. The Australian system shares many attributes with the US system, for example: the common law, a federal state structure, a dual role for the apex courts as both appellate and constitutional, concrete judicial review, and politicized judicial selection, amongst others. In this way, we could readily adopt the coding schema of a significant portion of the SCDB, adapting the details for the specifics of the Australian legal system. However, there are many areas in which direct comparison reveals contrasts, including: dissent in apex court decisions is less frequent in Australia; justices tend to write individual separate concurring and dissenting opinions; and the Court does not always sit en banc, instead frequently selecting panels to hear full court cases. Other differences include a greater number of jurisdictional possibilities for constitutional and statutory avenues for the apex court, the length of time for oral argument (which can range from hours to days), and the inclusion of the review of state court decisions on state law within the High Court’s jurisdictional authority. These differences do not necessarily imply any changes to the concepts of measurement, but simply require adaptations to the range of possibilities in the Australian case, for example, expansive code frames for state and local courts as well as administrative bodies (see normalizations on case source and administrative action), and areas of law and statute (see normalizations for legal issue and legal provisions respectively). There have been differences in the Australian literature on the High Court about how to code dissenting and concurring opinions (see Lynch, 2002, 2005; Groves & Smyth, 2004 for differing perspectives and a discussion of the issues). The HCDB coding specifies that a dissent is where a justice disagrees with the disposition of the case (i.e., the order). Where a justice agrees with the disposition but disagrees with the reasoning of a majority of justices that vote for a particular disposition, that opinion is coded as a concurring opinion. This coding choice maps onto that of the SCDB.

A key example where the HCDB deviates from the SCBD is with the inclusion of a subset of variables we have termed “gatekeeping variables.” This subset includes 15 variables that capture relevant information about the High Court of Australia’s grant or denial of applications to appeal to the apex court, termed an application for ‘special leave.’ Unlike the US Supreme Court, where the full bench of the Court votes on whether to grant ‘certiorari’ to an application to appeal, the High Court of Australia sits in panels of two or three justices (assigned by the Chief Justice), who determine whether leave to appeal will be granted or denied. In addition, until 2016, most applications for special leave were determined after a short oral argument where lawyers presented their arguments for the grant or denial of special leave to the panel of justices assigned to the case. As a result, we developed specialized coding schema for this unique feature of the High Court of Australia.

The unique institutional design of special leave in Australia also provided an opportunity for expansion on the initial database. The process of judicial agenda setting—both the special leave to appeal application process, and hearings that may follow, in Australia and the certiorari briefs, and their review process, used elsewhere—is an important avenue of research in judicial behavior (e.g., Boucher & Segal, 1990; Caldeira et al., 1999). This institutional design feature raises nuanced and complex questions about the role of individual justice preferences on institutional agenda setting, inter-justice dynamics on special leave panels, the role of the Chief Justice in controlling agenda setting when selecting special leave panels, and sub-majoritarian rules for agenda setting (Jacobi et al., 2023).

In our database, we link data on special leave decisions (in the public domain) by case and justice. Data which is collected and set for future public expansion are the transcripts of oral argument, linked by case, justice, and advocates, already partially analyzed in studies on the strategy of oral argument (Jacobi et al., 2023) and gendered interjections (Jacobi et al., 2024). Figure 1 gives the structure and linkage between the High Court database and its extensions.

Figure 1.

Diagram showing the linkage between datasets in the Australian High Court Database

In addition to the linked extensions relating to special leave, we have made several additional expansions upon the simplest form of the data, notably our associated dataset on justices’ biographical information. We link the core dataset which is the justice-centered High Court decisions and is structured as a line of data for each justice and for each case, 1995–2020. Further, we link a dataset recording the personal characteristics of the legal representation of each side of each case.

Technical Validation

To ensure the quality of data entries, we assessed intercoder reliability in the collection phase to ensure a clear understanding of the concepts being measured, and performed technical validations in the post-collection phase to clean unintentional mistakes from the data. All data in each dataset were hand coded. To ensure data quality and validity, each variable was coded by two coders independent of the project leads (i.e., the authors). We employed Cohen’s Kappa intercoder reliability rating to assess the validity of the coding (Cohen, 1960). The reliability rating was assigned on both the entire corpus of a single variable, and additionally on an annualized basis. Where the rating fell below 0.85, the variable was coded by a project lead and compared to each independent coder. In addition, a project lead independently coded and assessed a random sample of 10% of the entire corpus of variables with intercoder reliability ratings between 0.85–1.0.

After data collection, we performed simple tests to ensure that variables contained only values permitted by the coding frame and corrected where errors were found. We also labelled coded categorical data to aid analysis interpretation. These procedures were carried out in R, verified in Stata, and are included in our replication data within the included hcdb R package.

Australian Judicial Behavior in Comparative Perspective

In this section we analyze and discuss Australian judicial behavior in direct comparison with other countries. Our purpose is to illustrate the breadth of variables available across the database (justice_decision) and its extensions (justice_bio and special_leave) presented in Figure 1 and provide examples of how these interlinking datasets can be combined with other extant national databases to further research on comparative judicial behavior. Although the scope of our analysis is necessarily limited by the format of our article, we can arrive at several new conclusions which are helpful to scholars of comparative judicial behavior.

We first present comparative data on rates of judicial dissent. Comparative analysis of dissent in High Courts is important for two reasons: first, judicial dissent is indicative of the potential for ideological disagreement and is, thus, evidence in favor of the attitudinal model. Second, the rate of dissent is easier to calculate than ex ante ideology scores, the liberal/conservative coding of decision directions, and does not require justice level data (which facilitates ideal point analysis). It is, thus, more widely available to comparative scholars. However, the current empirical foundations of comparative dissent rely on data which is now more than 20 years old (Alarie & Green, 2017). Our analysis updates this literature with new data and largely confirms the statistical patterns found in previous studies, providing convergent validity and assurance to scholars that comparative empirical claims of the type “dissent is more common in Australia than in the UK” are well supported by modern data.

Next, using ex ante measures of judicial ideology and liberal-conservative decision direction, we provide the first direct comparison of ideological voting between the High Court of Australia and the United States Supreme Court. Australian legal academics have traditionally been resistant to the notion that the attitudinal model applies in Australia to the same degree as it does in the United States. One reason often given for the diminished role of judicial ideology in the High Court of Australia is that because Australia does not have a constitutional Bill of Rights, the policy-making function of the High Court has never been as explicit as the United States Supreme Court (Smyth, 2000). Sawer (1967) argues that the nature of the main issues considered by the High Court has traditionally differed from those of the US Supreme Court. In the United States, the cases involve fundamental liberty issues, as opposed to federal demarcation issues, which contribute to the politicisation of decisions. ⁵ This distinction, however, began to erode when Sir Anthony Mason was Chief Justice (1987–1995). Lindell (1998, p. 86) argues “perhaps the most important aspect of the Court’s work [in the Mason court was] …. its concern with protecting the individual against the abuse of public and private power”.

Dixon and Lau (2015) suggest that in Australia there is a distinction between legal (small ‘c’) and political (Capital ‘C’) conservatism. While Capital ‘C’ conservatism is marked by a commitment to originalism, Dixon and Lau (2015) argue that small ‘c’ conservatism is characterized by close adherence to precedent, a commitment to ensuring that the law-making role of the courts is exercised with restraint and that in cases of doubt that the courts defer to parliament. In the United States, legal and political conservatism is more closely related – i.e., political conservatives are more likely to adopt an originalist method to constitutional interpretation, while in Australia, there may be greater divergence. The distinction between legal and political conservatism in Australia complicates interpretation of the application of the attitudinal model in Australia and the United States.

Yet, irrespective of the exact version of attitudinal model advanced, there is much work by political scientists suggesting that there have long been ideological divisions on the High Court. The results from a series of scalogram studies of the High Court, covering the period from the mid-1930s until the mid-1970s – a period when the notion of strict and complete legalism espoused by Sir Owen Dixon had its strongest acceptance – found that the Latham, Dixon and Barwick courts all had liberal and conservative wings (see Blackshield, 1972, 1977, 1978; Douglas, 1969; Schubert, 1968, 1969). More recent work, including a study that employs an ex ante measure of judicial ideology based on newspaper articles prior to each justice taking their seat on the High Court, confirm that judicial ideology influences voting behavior on the High Court (Robinson et al., 2021).

Our findings build on this research to make it clear that ideological decision-making is indeed comparable between Australia and the United States. Although the relative lack of dissent in Australian cases constrains justices to a narrower band of voting preferences than in the United States, pre-appointment ideology and liberal voting correlate strongly and similarly in both courts.

Finally, we analyze judicial agenda setting across Australia, the United Kingdom, and the United States, where there are large differences in the rates of permission to appeal. Ultimately, we hope this overview of the data will encourage others to use the HCDB information for more in-depth comparative studies in the future. ⁶

Comparative Rates of Judicial Dissent

An important prerequisite of any measurement of judicial ideology is that justices disagree. Without disagreement, it is not possible to scale individual justices’ preferences on any dimension, be it left-right, pro- or anti-precedent, in favor of government institutions or individuals. Studies have compared dissent rates (Alarie & Green, 2017), using data collected on apex courts in several jurisdictions (Haynie et al., 2007). Alarie and Green (2017) analyzed rates of dissent in Australia (using earlier data, no longer publicly available), Canada, India, the United Kingdom (Law Lords) and the United States. They found that dissents were most likely in the US, followed by Australia, then Canada, the UK Law Lords, and finally India. ⁷ Our analysis replicates these findings using the HCDB and expands the analysis with data from the Philippines (Haynie et al., 2007), Israel (Weinshall & Epstein, 2020), and the UK Supreme Court, which succeeded the Law Lords as the highest court of appeal in the UK from 2009 (Hanretty, 2020b).

Our initial analysis of judicial dissent shows the spread of dissent rates across eight apex courts, presented in Figure 2. The analysis tentatively suggests two conclusions: ⁸ first, convergent validity of separate data collection efforts on dissent rates; and second, the importance of institutional differences—vis-à-vis cultural differences—in determining comparative rates of judicial dissent. First, there is convergence between separate data collection efforts in Australia and the United Kingdom. In the case of Australia, the dissent data we present is similar to the High Court data in Haynie et al. (2007), except our collection covers the period 1995–2020, while the previous analysis comprised cases from the period 1969–2003. The results are similar for both datasets in that Australia’s rate of dissent sits between the United States and Canada (for results using the Haynie et al. (2007) data, see Alarie & Green, 2017, p. 213). In the case of the UK Law Lords’ transition to the UK Supreme Court, dissent rates were also similar across both datasets. ⁹ This is not surprising as all twelve Law Lords were appointed as the Court’s first bench of Justices when the Supreme Court was established in 2009.

Figure 2.

Cases with one or more dissents in apex courts across seven countries

The second tentative finding concerns theoretical explanations for rates of dissent in different courts, and whether to attribute these differences to behavioral norms or institutional design. For example, the United States has very high rates of dissent in comparison with India, which may be reflective of a more combative political culture, or reflective of broader structural factors in institutional design. Alarie and Green (2017, pp. 211–46) argue that norms are an important explanatory factor for the rate of dissent, and that cultures of deference can cause a dissent rate to be higher or lower. They use as examples the difference between the US Supreme Court, which is described as an adversarial legal culture with high dissent, and the Indian Supreme Court, in which junior justices are expected to defer to senior justices.

However, there may be institutional factors which provide context for differing dissent rates beyond judicial culture. The US Supreme Court (as we show later) also has a very tight control over its docket, accepting only 4.5% of certiorari applications. On the other hand, Robinson (2013) shows that docket control in the Indian Supreme Court follows a very different institutional logic in which the status of cases filed for review may not be determined for several years. This is one example of an institutional or behavioral feature of these Courts which could partially explain differences in dissent. Further, the claim that norms are causal elements in judicial behavior is contested. Most rational choice accounts view norms as epiphenomenal—that is, they do not independently cause judicial behavior—but rather are byproducts of deeper structural forces (Epstein & Knight, 1998; Segal & Spaeth, 2002). Similarly, critical legal scholars see norms as power justification devices (Kennedy, 1997). ¹⁰ For these reasons, we suspect that India and Israel, which also hear a high number of cases in comparative perspective (Weinshall & Epstein, 2020), may owe just as much (if not more) to comparative differences in institutional design and workload than to cultural expectations or norms.

Attitudinalism in Australia and the United States

Segal and Cover (1989) tested the attitudinal model using ex ante measures of judicial ideology; their study is a vital foundational component of comparative judicial behavior, yet until recently there was limited comparative study on the topic (Weiden, 2011). In this section we present findings from the first such replication (Robinson et al., 2021) using the HCDB and comparing directly with up-to-date findings on the same measures from the United States Supreme Court, using Spaeth et al. (2024) and Epstein et al. (2022).

The findings here underscore that the attitudinal model applies in contexts without entrenched rights, such as Australia, which lacks a bill of rights (which is typically assumed to be an important cause of ideological disagreement). ¹¹ This suggests that judicial ideology may find expression within quite contrasting institutional designs. Indeed, examining judicial ideology in a parliamentary setting without constitutionally enshrined individual rights, we argue that judicial nominators (and their political persuasions) are often decisive in shaping the structure of ideological dispute on apex courts: the more partisan the appointment process, the more likely ideological expression among judges becomes (Robinson et al., 2021).

In both Australia and the US, the ideology score for each justice (JI) is measured by the Segal-Cover formula $JI=\left(Liberal-Conservative\right)/\left(Liberal+Moderate+Conservative\right)$ (Segal & Cover, 1989), normalised to produce an ideology score ranging from 0 (most conservative) to 1 (most liberal). The measure “liberal opinion” is the proportion of all opinions (including unanimous decisions) in which the justice voted for a liberal outcome.

In creating the Australian ideology scores, we adapt the definitions to include areas of law specific to Australian jurisprudence and code the following decision types as liberal (in the context of cases involving rights): pro-person accused or convicted of a crime; pro-civil liberties or civil rights claimant; pro-indigent; pro-rights claimant in a constitutional rights claim; pro-privacy; and pro-Indigenous rights. A full list of coding rules for determining the ideological direction of decisions is available in the justice level codebook (Section 127).

The findings are given in Figure 3. Each graph describes a similar pattern: as liberal (cf. conservative) ideology increases, so too does the proportion of votes in which each justice votes in a liberal direction. To be specific, both scatter plots describe strong positive correlations (p < 0.001), with $\rho =0.68$ in the United States Supreme Court, and $\rho =0.61$ in the Australian High Court. Our measure of the proportion of liberal votes includes both unanimous and non-unanimous cases. We believe this is important for comparison between the two courts, since the likelihood of dissent in the US Supreme Court is higher, and so there is greater scope for visible ideological disagreement. Of course, correlation shows association between two variables, but not the magnitude of a relationship. The slope of a simple linear regression shows that the strength of the relationship between pre-appointment ideology and career-long voting patterns is twice as strong in the United States ( $\beta =0.31$ ) as it is in Australia ( $\beta =0.15$ ).

Figure 3.

Scatter plots of pre-appointment liberal ideology and liberal decision-making in the High Court of Australia and US Supreme Court

In this first direct comparison between the US and another jurisdiction that follows the Segal-Cover methodology, we provide suggestive evidence that the strength of attitudinal voting is structurally related to the rate of dissent. It highlights the analytic depth of our data, and demonstrates the principle of harmonized data design, which enables precise and direct comparison beyond what is typically collected and analyzed in jurisdictions outside of the United States.

Comparing Judicial Workload Management

In this final analysis section, we compare how Courts in Australia, the United Kingdom (Hanretty, 2020b), and the United States (Bonica et al., 2025) control their dockets, and consequently, their agendas. A Court’s ability to control its own agenda has political and legal implications. In deciding which cases to hear (and which not to), the Court has significant scope to impact both law and society through its interpretation of constitutions, statutes, and common law. These decisions are unlikely to be purely apolitical, especially if partisan politics has inputs into judicial selection.

The US Supreme Court maintains strong agenda control; under 5% of cases since 2000 were granted other than through a discretionary writ of certiorari (Spaeth et al., 2024). Meanwhile, in Australia, the assertion of the special leave requirement has slowly consolidated the High Court’s control of its docket, with approximately 84% of cases appearing via the Court’s discretionary docket. In the UK, Hanretty (2020a, p. 63) shows that the UK Supreme Court has still less control, with approximately 70% of cases granted permission to appeal by the court itself, with the other cases mandated either by lower courts or as-of-right appeal. A similar pattern applies when examining grant rates for applications to appeal at the court’s discretion. Figure 4 shows that the United States Supreme Court accepts only 5% of cert applications, the High Court of Australia accepts 11%, while the United Kingdom Supreme Court accepts 33%.

Figure 4.

Grants of leave to appeal (certiorari) in Australia, the United Kingdom and United States

These figures suggest that the three courts may be placed on a spectrum from relatively little docket control to extremely tight, but it is important to place these considerations in terms of overall workload, the demand for appeals from lower courts, and the institutional rules which determine grants of permissions to appeal. When considering overall caseload, the US Supreme Court still grants certiorari ¹² to roughly three times as many cases as the High Court of Australia—although notably, the hearings last days, rather than minutes, as in the US (Jacobi et al., 2023), and nearly two and a half times as many cases as the UK Supreme Court. When considering the smaller number of cases, the greater number of justices, and the tendency of the UK Supreme Court to hear cases as specialist panels, the UKSC’s high acceptance rate may not reflect a lack of agenda control, simply that they receive fewer applications relative to their capacity to hear cases.

Further, considering the rules by which permission to appeal is granted, there appear to be soft institutional arrangements that are calibrated to the level of demand for appeals to each Court. In the US Supreme Court, receiving an average 3,547 paid certiorari applications per year, ¹³ and more than twice as many unpaid petitions, ¹⁴ a rule of four requires a near majority of justices of the bench to consider a case worthy of a full hearing.

In Australia, the High Court receives an average of 560 applications per year. Given that most special leave applications are heard by two justices (on a seven-justice bench) and that special leave panels decide unanimously, this effectively becomes a rule of two.

Meanwhile in the United Kingdom, where the Court receives an average of 336 applications per year, the rule is closer to a rule of one; the rest of the panel will normally acquiesce if a single judge wants to take a closer look at a case, and first opinion is given to the junior justice sitting on an appeal committee (Darbyshire, 2011, p. 440). In deciding cases in this manner, there is no suggestion that the UK Supreme Court (nor the Law Lords before them) are over-burdened by work (Darbyshire, 2011, p. 439). We might argue that these procedural dynamics are norm-driven, but it appears to be better explained by more prosaic organizational constraints: courts see as many cases as they can thoroughly process in a year.

In sum, our analysis of agenda setting highlights the extensibility of our dataset, which is a key principle behind our database design. Further, the above analyses on dissent and ideology point to the possibility that structural relationships between agenda setting and judicial ideology are valuable topics for future research. We do not offer conclusive theoretical or empirical findings here, but we do demonstrate that our dataset in combination with the latest data across jurisdictions makes this area of research far more accessible than ever before.

Guidelines for Use

Ultimately, we do not wish to be unduly prescriptive about how scholars use our data, as there are many justifiable analyses, most of which are beyond our imagination. Here, we elaborate briefly on three broad guidelines which should help users avoid the pitfalls of using these data in a comparative study. The guidelines are as follows: read the codebooks, understand the institution, and appreciate the limitations of the comparative method.

The codebooks provided with the data provide exhaustive detail on the data’s structure. In Figure 5, we detail one sample codebook entry which gives a description of the specific statute that formed the basis of a decision in the High Court. The Variable Label heading refers to the column header in the final dataset, while the Normalizations heading (varLegalProvision) refers to the list of all possible values taken by the variable (with the number of distinct values in parentheses). These normalizations are frequently shared between variables, as in the case of legal provisions where coders have up to three windows to record relevant statutes (lawSupp1, lawSupp2, and lawSupp3). The legal provisions in the normalization are organised hierarchically according to another normalization (varLawArea), which groups ten broad areas of law on the topic from federal constitutional law to state educational statutes. A close reading of the codebooks should therefore allow scholars to feel confident that they can extract the precise information which is most relevant to them.

Figure 5.

Example of a codebook entry from the dataset on justice decisions in the High Court

Second, it is important that for in-depth comparative studies that researchers examine the institutional and historical context of the Australian High Court. We have given a very brief introduction to the Court in this article, but scholars may find it useful to consult the many excellent reference sources for a more thorough grounding. ¹⁵

Finally, scholars should be aware that the study of cross-national judicial behaviour is challenged by inherent problems in the comparative method (Collier, 1993). First, we have a limited number of countries. The above analysis demonstrates that we have relatively few comparable quantitative datasets available to us, and we have many more variables than countries on which we might compare. Second, the countries we do have are not representative of apex courts around the world and are heavily biased toward English-speaking, common law, or global north countries. Until there exist a greater number of countries, particularly in the global south, our ability to generalise will be circumscribed. Third, concepts such as dissent or agenda control may shift substantially across legal contexts, which is why institutional and historical understanding is indispensable. Finally, any claim of strict causality from quantitative comparative analysis is tentative and abductive in nature (Spirling & Stewart, 2025). However, this does not prevent researchers from theory building, taxonomizing, or asking causal questions, all of which contribute enormously to improving our understanding of behavior in apex courts.

Conclusion

In this article, we have introduced the HCDB and shown that the principles behind its compilation are suitable to direct cross-jurisdictional comparison. While cross-national efforts to improve data infrastructure in comparative judicial behavior have made substantial progr in recent years, there is much still to do. The empirical literature on judicial behavior in Australia has been described as “fairly thin” (Smyth, 2021, p. 108). This applies a fortiori to studies of comparative judicial behavior that include the High Court of Australia. The result is that many of the methodological advances, as well as richness of constructs that are now features of much empirical judicial work in other jurisdictions are yet to be fully reflected in research on the High Court of Australia (Smyth, 2024). With this public data resource, we provide a means of conducting rigorous empirical research on the High Court of Australia.

We have also illustrated here several different possible lines of inquiry to which such data can be put, allowing not only analysis of the High Court of Australia, but providing the tools for comparative inquiry. These include addressing fundamental questions about judicial behavior, judicial agenda setting, judicial workload, and others. We made the choice to err on the side of over-inclusion of variables—contrary to some recommendations—to enable the broadest possible range of inquiries for others to pursue using our tools.

Finally, this paper has demonstrated the utility of our database design and has focused on facilitation of direct comparison through codebook harmonization, wherever possible. Future work may see comparative scholars take on this challenge collaboratively. In political science more broadly, the first task for comparative researchers when collecting data is to create a collaborative and harmonized codebook, to aid cross-national efforts at data collection. Examples of this tendency in comparative politics abound: The Comparative Manifesto Project, the Chapel Hill Expert Survey, Varieties of Democracy or V-Dem, the Comparative Study of Electoral Systems, the Global Barometer Surveys and many others. This should also be the case in comparative judicial behavior. The efforts of the National High Courts Database are now more than two decades in the past, and there is much new cross-national data that has not yet been harmonized. We believe that renewed attention to a cross-national codebook is necessary to make harmonization ever simpler.