A Rose by Any Other Name: Understanding Judicial Decisions that Do Not Cite Precedent

• Author: Kawin Ethayarajh,Albert H. Yoon,Andrew Green
• Date: 01 September 2018
• DOI: http://doi.org/10.1111/jels.12186
• Published date: 01 September 2018

Journal of Empirical Legal Studies

Volume 15, Issue 3, 563–596, September 2018

A Rose by Any Other Name:

Understanding Judicial Decisions that Do

Not Cite Precedent

Kawin Ethayarajh, Andrew Green, and Albert H. Yoon*

In common-law countries, legal precedent serves as a foundation of judicial opinions.

Judges cite precedent to explain their decision, and it is this use of precedent that threads

one decision to another. The Supreme Court in India stands in contrast to its counterparts

in other countries in that it annually decides not dozens, but thousands, of cases. Perhaps

unsurprisingly, nearly half the Court’s decisions do not cite any precedent at all. This arti-

cle examines this phenomenon, specifically how it affects judges’ commitment to the com-

mon law, in substance if not in form. Examining every Court decision for the period

1950–2010, we textually analyze the opinions using machine learning to determine what

connection, if any, exists between cases. We find that it is possible to accurately model how

the Court cites to existing precedent and that even for decisions without any citations,

there is almost always at least one prior decision the Court could have cited. Our finding

suggest that time and resource demands are primarily responsible for the failure to cite

relevant precedent, but that the Court acts efficiently, given the constraints placed on it,

in deciding in which decisions to include precedent. This research, however, leaves unan-

swered whether the Court provides sufficient guidance to lower courts.

I. INTRODUCTION

Faith in the common law is premised on consistency and transparency: consistency, in

that courts decide similar cases similarly; transparency, in that courts explain how they

reached their decisions. To achieve these goals, courts cite precedent to show how prior

decisions inform their reasoning. Courts can break with their earlier decisions, but are

expected to be explicit in their reasons for doing so. Given the central role that prece-

dent plays in the common law, can the common law function without such citation? Can

consistency be maintained in other ways that allow the common law to fulfill its role in

*Address correspondence to Albert Yoon; email: albert.yoon@utoronto.ca. Ethayarajh is at the University of

Toronto Department of Computer Science; Green and Yoon are at the University of Toronto Faculty of Law.

We thank Ben Alarie, Anita Anand, Ronen Avraham, Shubha Ghosh, Ed Iacobucci, Jonathan Nash, and

Anthony Niblett for their helpful comments. We also thank workshop participants at the Tel-Aviv-Toronto-Sienna

and Canadian Law & Economics conferences. Misha Boutillier provided excellent research assistance. All remain-

ing errors are our own.

563

different contexts? In many common-law jurisdictions, where courts write expansive opin-

ions, we may take precedent for granted. However, it is worth noting that this tradition of

stare decisis was not always so. Even at the high-court level, we can see instances where

judges do not cite prior cases. In the 1800s, less than 20 percent of U.S. Supreme Court

cases cited at least one prior decision, although by the early 1900s, over 90 percent of

Supreme Court decisions contained at least one citation (Fowler & Jeon 2008).

The Supreme Court of India (SCI) operates in contrast to its counterparts in the

United States, England, and most other common-law jurisdictions. The SCI consists of up

to 31 judges who typically sit in panels of two. It chooses to hear an expansive caseload

that routinely numbers into the thousands annually, in contrast to supreme courts in

most other countries that routinely hear fewer than 100 cases each year. Over time, the

Court has taken on an increased role in providing access to justice, in part through

expanding the opportunity for individuals to bring public interest actions directly before

it. At the same time, the Court has undertaken functions that would fall to the executive

in many countries, such as directing education and environmental policy (Robinson

2013b). However, these undertakings may come at a cost to the system of precedent

(Robinson 2013a). The Court cites precedent in a manner comparable to the U.S. or

Canadian Supreme Courts only in a small percentage of cases (Green & Yoon 2017). In a

large fraction of cases—40 percent, with some years exceeding 50 percent—the Court

does not cite any precedent at all.

1

Hence our question: How can a high court that does not cite its own prior deci-

sions in a substantial fraction of cases ensure the rule of law and the development of the

common law? An easy but troubling explanation is that the SCI—given its prodigious

caseload—may be able to focus its energies only on the few cases it deems as truly impor-

tant, with the remainder simply being decided on an ad hoc basis. The judges on these

other cases may be simply too overwhelmed by their caseload to search for relevant prior

cases and ensure consistency in the case law.

A more sanguine explanation is that judges find connections and maintain consis-

tency in subtler ways, such as through informal communications between judges or by

ignoring recent case law in exchange for relying on principles in older cases that they

remember from law school or practice. Stated another way, the SCI may reach the appro-

priate legal result, albeit without the transparency that including precedent provides.

These alternative forms of attaining consistency may be efficient in the sense of allowing

judges to reduce search and decision costs, but may impose negative externalities on the

legal system as a whole. These externalities include slower adaptation of the law, bias in

the development of the law based on the composition of the informal networks, and

inability of lower courts and litigants to see the consistency given the lack of citation.

This article examines what courts do when deciding cases without citing to prior deci-

sions. Considerable empirical work has been done on the U.S. Supreme Court and, to a

lesser extent, courts with similar designs such as the Supreme Courts in Canada, the United

Kingdom, and Israel. We focus on the ISC, which has been surprisingly ignored, given that

1

See Figure 2.

564 Ethayarajh et al.

India represents the world’s largest democracy. It is important to understand its functioning

in its own right, given that India has the world’s second largest population and seventh larg-

est economy (World Bank 2016). The SCI also provides us with an opportunity to examine

how an alternative design of a high court functions and what it can tell us about how the

design of courts in countries such as Canada and the United States may be improved.

The article proceeds as follows: Section provides background on the theory of how

courts use precedent, along with an outline of the empirical work on precedent. Section

offers a brief overview of the SCI before discussing our data. We have created a database

of the full text of all decisions

2

of the SCI since its inception in 1950 through 2010. We

therefore have information about each case, including whether or not it cited precedents

and, if so, which cases were cited. Section IV sets out our methodology. Our first step was

to examine decisions that did cite prior cases. Building on network analysis, we used

machine learning—specifically, a Support Vector Machine (SVM) with a radial basis func-

tion (RBF) kernel—to determine whether a connection exists between the cases that do

cite precedent and all the decisions that could have been cited by those cases. We then

use the resulting model of which cases judges cite to examine the decisions where the

Court does not cite any prior cases. The model allows us to search for connections

between decisions that do not contain any citations and the stock of all decisions to

which that case could have cited—that is, we can predict which cases the Court would

have cited to (but did not) if it had been acting consistently with the cases when it did

cite. Section V presents our findings. We find that for decisions where it does not cite to

precedent, there is almost always at least one prior decision to which the SCI would have

cited had it behaved as it did with the cases when it did cite. The judges’ reluctance to

cite precedent in a large fraction of cases appears to be partly based on high search costs

relative to the importance or function of the particular decision. We discuss major impli-

cations in Section VI. Section VII concludes.

II. UNDERSTANDING PRECEDENT

The common law is built on the system of precedents. The pinnacle in most countries is

the supreme court: its decisions bind all lower courts. When deciding cases, the supreme

court is, in theory, at least partially constrained by its own earlier decisions. However, it

may diverge from a prior decision by distinguishing the current case on its facts or over-

turning the prior decision altogether (Knight & Epstein 1996). The extent to which a

supreme court feels bound by its own prior decisions may vary by country, by area of law,

or over time.

Precedents provide two broad benefits. First, they provide legal principles, which

aid in identifying relevant facts and the legal consequences of those facts (Black &

Spriggs 2013). The stock of prior decisions contains the aggregation of the work of suc-

cessive courts in addressing particular issues (Landes & Posner 1976:250; see also Black &

2

This excludes decisions that resulted in no written order.

Judicial Decisions that Do Not Cite Precedent 565