• Introduction

  Such legal research employs an empirical method to draw inferences from observations of phenomena extrinsic to the researcher. Putting it simply, legal researchers often collect and then analyse material (data) that they have read, heard or watched and subsequently make claims about how what they have learned may apply in similar situations that they have not observed (by inference).¹x For insight into the extent to which legal researchers undertake empirical research and the lack of clarity around empirical methods in law see Epstein, L. and King, G. (2002) ‘The Rules of Inference’ Vol. 69 No. 1 The University of Chicago Law Review 1-133 at 3-6, Part I. One such form of empirical method is the case study, a methodological term which has been used by some researchers to describe studies that employ a combination of data sources to derive in-depth insight into a particular situation, by others to denote a particular ideological approach to research recognizing that the study is situated within its real-world context.²x Yin, R. K. (2014) Case Study Research Design and Methods (5th edn.) Sage Publications, 12-14. It is consequently a flexible definition encompassing approaches to the data and the stance of the researcher,³x See Hamel, J. with Dufour, S. and Fortin, D. (1993) Qualitative Research Methods Volume 2, Sage Publications, ch 1. but its malleability has led some researchers incorrectly to stretch the term to encompass any study that focuses on one or a restricted number of situations.⁴x See Gerring, J. Case Study Research: Principles and Practices, Cambridge, Cambridge University Press, 2007 at 6. For a further discussion see Gerring, J. ‘What is a case study and what is it good for?’ (2004) Vol. 98 No 2 American Political Science Review 341-354. This looseness in definition in a legal context may perhaps be linked to confusion as between teaching and research case studies; some traditions in legal education employ a teaching method known also as ‘case study method’ which operates quite differently from its research counterpart. For a discussion of the differences between teaching and research case studies see Yin, 2014, 20 and for a discussion of teaching case studies see Ellet, W., (2007) The Case Study Handbook: How to Read, Discuss, and Write Persuasively About Cases, Boston MA, Harvard Business Review Press; Garvin, D. A. (2003) ‘Making the Case: Professional Education for the World of Practice.’ (Sept–Oct) Harvard Magazine 56-65. This has resulted in concerns that the definition has been co-opted as a means to explain any small n⁵x ‘n’ (number) is used to denote the number of observations in the study, N is used to describe the total number within the population when n denotes the sample observed. empirical study that has a focus on a particular subject, time-frame or location, and further that this has led to poor quality empirical research in law.⁶x For a discussion of the state of empirical research in law see Epstein and King, 2002. This is perhaps unsurprising, as law programmes tend to be very strong at teaching lawyers how to source, interrogate and then draw valid inferences from legal data sources such as cases and legislation, but less adept in the context of other types of data (for example survey data, interviews, non-legal documents and/or observation).⁷x See Webley, L. (2010) ‘Part III Doing Empirical Legal Studies Research Chapter 38 - Qualitative Approaches to ELS’ in Cane, P. and Kritzer, H. (eds.) Oxford Handbook of Empirical Legal Studies, Oxford, Oxford University Press. Case study method usually involves an array of research methods to generate a spectrum of numerical and non-numerical data that when triangulated provide a means through which to draw robust, reliable, valid inferences about law in the real world.⁸x For a discussion about the differences between numerical (quantitative) and non-numerical (qualitative) data see Webley, id; Epstein and King, 2002, at 2-3; King,G., Keohane, R. O., and Verba, S., Designing Social Inquiry: Scientific Inference in Qualitative Research, Princeton NJ, Princeton University Press, 1994 at 6. It is relatively underused in empirical legal research.
  This article aims to make a contribution to those new to the case study method. It will examine the purpose of and why one may wish to undertake a case study, and work through the key elements of case study method including the main assumptions and theoretical underpinnings of this method. It will then turn to the importance of research design, including the crucial roles of the academic literature review, the research question and the use of rival theories to develop hypotheses in case study method. It will touch upon the relevance of identifying the observable implications of those hypotheses, and thus the selection of data sources and modes of analysis to allow for valid analytical inferences to be drawn in respect of them. In doing so it will consider, in brief, the importance of case study selection and variations like single or multi case approaches. Finally, it will conclude with some thoughts about the strengths and weaknesses associated with undertaking research via a case study method. It will address frequent stumbling blocks encountered by researchers, as well as ways so as to militate against common problems that researchers encounter. The discussion is necessarily cursory given the length of this article, but the footnotes provide much more detailed sources of guidance on each of the points raised here. This article is an introduction to a case study method rather than an analytical work on the method.

• 1. Case Study Method: Purpose of a Case Study, Why Undertake One?

  Case study method falls within the social science discipline and as such has scientific underpinnings. The case study examines phenomena in context, where context and findings cannot be separated. Case study design is also sometimes used to investigate how actors consider, interpret and understand phenomena (e.g., law, procedure, policy) and therefore allows the researcher to study perceptions of processes and how they influence behaviour, for example to understand judges’ sentencing choices in a Dutch police court.⁹x Mascini, P., van Oorschot, I., Weenink, D. and Schippers, G., (2016) ‘Understanding judges’ choices of sentence types as interpretative work: An explorative study in a Dutch police court’, (37) (1) Recht der Werkelijkheid 32-49. This may help to understand how laws are understood, and how and why they are applied and misapplied, subverted, complied with or rejected. This can flow back into the legal and policy making processes, court procedure, sentencing, punishment, diversion of offenders etc., and may have a high impact as a result. The conditions precedent for case study method have been succinctly explained by Yin as follows:

  ‘doing a case study would be the preferred method, compared to the others, in situations when (1) the main research questions are “how” or “why” questions; (2) a researcher has little or no control over behavioural events; and (3) the focus of study is a contemporary (as opposed to entirely historical) phenomenon.’¹⁰x Yin, 2014: xxxi and further 16-17.

  The key points to note here are that a case study is a real-world in-depth investigation of a current complex phenomenon. The research will take place in situ (rather than in the library or moot court room) where the researcher cannot control the behaviour of research participants.

  The purpose of the study is to learn how or why something happens or is the way it is, and this is achieved by collecting and triangulating a range of data sources to test or explore hypotheses.¹¹x Triangulation is the term used to explain that a research question is considered from as many different standpoints as possible, using as many different data types as possible to permit a holistic examination of the question to see which explanations, if any, remain consistent across all data sources. It caters for a wide range of modes of enquiry: the investigation may be exploratory (explore why or how something is the way it is), descriptive (describe why or how something is the way it is) or explanatory (determine which of a range of rival hypotheses, theories etc. explain why or how X is the way it is).¹²x Yin, 2014: 5-6. Some categorise case studies as those designed to be theory orientated, and those designed to be practice orientated.¹³x See Dul, J. and Hak, T. (2008) Case Study Methodology in Business Research, Oxford: Elsevier 8-11, 30-59. Thereafter the design scope is very broad; the data collected may be qualitative and/or quantitative, collected via a variety of methods, and the case study may be a single case or be made up of a small number of cases. The breadth of data collected may be illustrated by Latour’s ethnography of the Conseil d’Etat in France, which studied the connections between human and non-human actors to explore their relationship with ‘the legal’ and ‘the Law’ is assembled in that court context.¹⁴x Latour, B. (2010) The Making of Law: An Ethnography of the Conseil D’Etat, Cambridge: Polity Press. Case study method is a way of thinking about research and a process through which one seeks to produce reliable, fair findings. It can provide deep insight into a particular situation, whether particular in time, in location or in subject-matter.¹⁵x For a discussion of ethnomethodological aims to study practical life as experienced in context as an end in itself, as experience is subjective and situational, see Small, M.L. ‘‘How many cases do I need?’ On science and the logic of case selection in field-based research’ (2009) Vol. 10 (1) Ethnography 5, 18. It may allow for transferable findings in respect of the theoretical propositions/hypotheses being examined if not to a population as would often be the situation in much quantitative research.¹⁶x For greater insight on this point see Lipset, S. M., Trow, M. and Coleman, J.S. (1956) Union Democracy: The Internal Politics of the International Typographical Union, New York: New York Free Press at 419-420; Yin, 2014, 21. For a discussion of the problems inherent in aping quantitative terminology in qualitative work see, Small, 2009, 10, and at 19 for further reading on the logic of case study selection and further reading on extended case method. It aims to examine rival hypotheses, propositions, potential explanations previously advanced (exploratory study), or to test findings from a previous case study examining similar phenomena in a new instance (a replication or confirmation study).¹⁷x Gerring, 2007, 346.

  As described so far it is a research method that appears to have a lot in common with experiments and tests of statistical significance. But case study method differs markedly from a big data survey or double-blind experiment in that it seeks explicitly a phenomenon in its natural environment and (in most instances) without means to control for variables, including the behaviour of any participants.¹⁸x Although note that there are some scholars who believe that case study method can include elements of experimental testing, for example, Gerring, J. and McDermott, R. (2007) ‘An Experimental Template for Case Study Research’ Vol. 51 No. 3 American Journal of Political Science 688-701. One such study in law that has been described by some, if not by the researchers themselves, as a case study did include an experimental design within the battery of methods employed see: Moorhead, R., Sherr, A., Webley, L., Rogers, S., Sherr, L., Paterson, A. & Domberger, S. (2001) Quality and Cost: Final Report on the Contracting of Civil Non-Family Advice and Assistance Pilot (Norwich: The Stationery Office). Experiments aim to control some factors so as to test hypotheses under different conditions, quantitative studies attempt to control for environmental factors through sampling techniques and data collection instrument design so as to minimise their biasing effects, but case study method does not involve control of the environment, or control for the environment, instead it aims to harness context and work within it. It examines in great detail one situation (referred to as a case or unit) or a very small number of situations, to use context as a means to particularise the findings. It also seeks to explain which elements of context may mean that some of the findings are applicable to other situations and if so under what conditions. A case study tells the researcher about the case and the extent to which previous explanations are sustained, in some instances it may also allow the researcher to make claims that some of the findings can be applied to another case or cases too, although this is heavily dependent on the research design and its execution.¹⁹x Campbell, D.T. Foreword in Yin, 2014 xviii. But it is rarely, if ever, a method that can be used by one to want to make universal claims. A case is not a proxy for a sample of a population in a survey, for example, it is a study of a phenomenon in itself rather than a means through which to view the whole world. Having said that, samples can be used to help select cases in a sound manner.²⁰x Seawright, J. and Gerring, J. (2008) ‘Case Selection Techniques in Case Study Research: A Menu of Qualitative and Quantitative Options’ Vol. 61 No. 2 Political Research Quarterly 294-308.

  Case studies are only one of a number of ways to undertake socio-legal or criminological research and it is important to give proper consideration to the full range of research methods prior to making a final decision to adopt a case study method.²¹x Yin, 2014: chapter 1. It may be better to employ a different one: legal history; doctrinal legal study (legal cases, legislation, regulatory documents); a policy study (policy documents, communiqués etc.); a statistical analysis (an analysis of the number of different types of legal cases that go before the courts, their key features and what role these play in chances of success for the plaintiff); a large-scale survey; stand-alone interviews; or an experiment in a simulated setting (asking lawyers to read through some scenarios and explain what advice they would give to a client in those situations). But a case study could employ a number of these methods in combination, so how then does one determine whether case study method is right for one’s study? It will largely depend on the nature of the research question to be answered and one’s appetite for undertaking in-depth research aimed at achieving thick description (detailed description of how or why something is as it is)²²x For a discussion see: Ryle, G. (1949). The Concept of Mind. London: Hutchinson; Lincoln, Y.S. and Guba, E.G. (1985). Naturalistic Inquiry. Newbury Park, CA: Sage Publications; Holloway, I. (1997). Basic Concepts for Qualitative Research. London: Blackwell Science. and/or triangulated findings derived from a range of data sources that develop a new theory or test existing rival theories. It is an intensive study, it requires extremely good planning and design and a robust approach to data analysis too.

• 2. Case Study Method: Research Design

  Research design is of paramount importance in achieving a successful case study, especially so given that the study focuses on one or a small number of situations and the researcher’s in-depth knowledge of and immersion within that case may lead more readily to confirmation bias than in some other forms of study.²³x Confirmation or interpretive bias of data is something we all have grapple with, as the natural human tendency is to place more weight on evidence that confirms our view than on evidence that contradicts it. Strong research design can assist with counter-balancing this to some extent, including the transformation of any expected finding into a hypothesis that one then seeks to falsify rather than to confirm. Research design begins with the choice of research topic, usually then followed by a review of any relevant academic literatures (perhaps beyond the boundaries of one’s own discipline, for example sociology, criminology, political science) to determine an appropriate research question, noting all possible answers to the research question that are posited in the literature.²⁴x See Yin, 2014, chapter 1 for more information on the role of the literature review. During this iterative phase the research question will be further refined, so that it may be articulated with precision, which is particularly important for much case study research as the link between the question and the chosen case or cases is usually explicit and explained, consequently a clear research question is considered by many to be an essential starting point to aid the selection of cases to be examined. The rival hypotheses, theories or propositions that may answer the question should, normally, be similarly delineated and clarified, those that remain plausible answers to the research question should be retained and be supplemented with any others that the researcher considers to be alternative viable explanations. Other approaches may be used that are more inductive than deductive, as in the case of many ethnographic case studies such as Latour’s, for example. This phase is an intellectually demanding one, but it sets the foundation for a strong study that is easier to execute at the point of data collection. The literature review also helps to ensure one is up-to-date, that one does not make the same mistakes that earlier researchers have reported as hazards, and to add a theoretical depth to one’s study that aids sophisticated analysis. It may also help to identify useful data collection methods and instruments too. And so time spent on the literature review may be very profitable.
  So far the discussion has been very general and therefore a little abstract. At this stage it may help to consider a hypothetical research proposal for case study research and work though it as the article progresses. The researcher in our hypothetical scenario is interested in undertaking research on recent reforms to the use of family mediation in the family justice system in England and Wales. She knows that it is now compulsory in most instances for the person who is initiating any court proceedings in a divorce to have participated in at least one mediation information and advice session with the aim of negotiating an outcome in relation to children, money and/or property prior to initiating proceedings in court. She is clear on the law and the procedural issues but not clear on how effective have been these changes, and this is her broad area of interest. After completing her literature review she understands that the key aims of the reforms were to reduce the number of cases going to court by increasing the number of cases that result in negotiated agreements between the divorcing spouses and in doing so to reduce the cost and the time involved in reaching outcomes in divorce cases, reduce the need for people to use lawyers in the negotiating process, to reduce acrimony between the divorcing spouses and to reduce the negative effects on children. Further, the reforms were intended to promote more durable outcomes between divorcing spouses that could be renegotiated effectively if the arrangements for the children needed to be updated to meet changing circumstances. But the researcher still needs to work these insights into a research question before making a final decision on whether a case study is the best method by which to conduct the research. The next sub-section will consider the framing of the research question, and will include examples of how our researcher may draft her question to maximise her chances of undertaking a great study on her area of interest.

  A. The Research Question

  The process of defining the research question may be a painful, frustrating one but it could also be creative too. It may be necessary to spend a considerable period of time reading the literature so as to narrow down the research topic or statement to a manageable, novel and/or important and scholarly question.²⁵x For more assistance with legal research questions refer to Epstein, L. (1995) Studying Law and the Courts in Lee Epstein (ed) Contemplating Courts, Cong Q, 1, 3-5. Some argue it should also seek to address a real world problem, although that is a controversial component and suggests that knowledge for knowledge’s sake is not a legitimate aim.²⁶x See King, Keohane and Verba, 1994, 15. The development of a research topic into a research question with reference to the academic literature is sometimes described as the phase in which the researcher has a conversation or dialogue with the literature. This dialogue grounds the study, it also informs the study design, including the case selection and data to be collected.
  The research question (a statement that ends with a question mark) is made up of two key elements: its substance, the topic or issue that you wish to address and the form of the question ‘who, what, where, how, why’.²⁷x See Yin, 2014, at 11, and see further Campbell, J.P., Daft, R.L., Hulin, C. L. (1982) What to Study: Generating and Developing Research Questions (Studying Organizations), Sage Publications, for further thoughts on research questions The substance of the research question is not simply the topic but the specifics of the topic – is your study to be a contemporary one or a historical one? In what context are you operating? What precisely are you endeavouring to study? The form of the question is also important: as indicated previously, case study method is considered to be better suited to research questions framed in ‘how’ or ‘why’ terms. Single case studies are considered to be an excellent means by which to uncover and understand the processes or mechanisms that influence particular variables (known as process tracing²⁸x On process tracing see: Collier, D. ‘Understanding Process Tracing’ Vol. 44 Political Science and Politics 823-830 and George, A. L. and Bennett, A. Case Studies and Theory Development in the Social Sciences, Cambridge, MA, MIT Press, 2005. For an excellent insight into how this has been used in a legal and policy context with reference to changes in Georgia’s tax laws see Ulriksen, M.S, and Dadalauri, N. ‘Single Studies and Theory-testing: The Knots and Dots of the Process-tracing Method’ (2014) International Journal of Social Research Methodology 1- 17. ), why or how different variables are related to each other, for example what influences legislative change or policy formation on a given topic in a given country at a particular time. They are also a sophisticated means through which to test empirically and deductively the congruence of rival explanations (theories or propositions), ‘to what extent’ or ‘how’ and ‘why’ different theories are borne out by the data.²⁹x See Blatter, J. and Haverland, M. Designing Case Studies, Basingstoke, Palgrave MacMillan, 2012 at 145 who consider process tracing involves inductive reasoning to build theory and congruence testing involves deductive reasoning to test theories. The form of the initial research question can confound some researchers who initially phrase their question as a ‘what’ question and as a result unnecessarily rule out case study method. Questions can often be reframed, for example: ‘what have the prosecutorial authorities in England and Wales done to integrate victims of domestic violence into the criminal justice process?’ may be rephrased as ‘how have the prosecutorial authorities in England and Wales integrated victims of domestic violence into the criminal justice process?’.
  At this stage our hypothetical researcher is faced with some choices: should she consider ‘how have the reforms to family mediation used in the divorce context affected the durability and suitability of post-divorce arrangements in England and Wales?’ This would focus on the agreements whether agreed or adjudicated, their longevity, the extent to which they could be made to work after the divorce and how any amendments to arrangements were sought. Alternatively, she could ask ‘how have the reforms to family mediation affected the way in which divorces are conducted in England and Wales?’ This would examine the steps people took so as to get divorced but may also consider the divorcing couples’ perceptions about the process to assist with examining the policy to reduce acrimony, it could also address how much time and money they spent in the process and it could also elicit data on how constructive was their relationship and negotiations subsequent to the initial agreement or adjudication. It could also bring in the role of lawyers and/or consider the children’s experience of the divorce process too. The research design would then follow the focus of the question.
  Case studies are much easier to design when the research question is expressed clearly, the theory is used to provide possible answers that may be explored or tested and the boundaries of the study are articulated. Some people find it helps to break down a draft research question into its substance and its form, and describe the purpose of the study in a couple of lines too,³⁰x Epstein and King suggest a range of possible purposes, at 59, including: to explore something that has not previously been studied; to attempt to settle a debate that has been ongoing within the literature; to examine a well-considered question but in a new way; to collect and analyse new data to seek to confirm or refute previous findings; to analyse an existing data set in a new or better way to seek to confirm or refute previous findings or to develop new ones. and then compare the extent to which all three are congruent and precise before moving on to the next phase of the design process. We shall consider the important role of theory in the next sub-section.

  B. The Theory

  Case studies afford the opportunity to observe a sequence of events or factors, to evaluate which produce an outcome and why,³¹x Peters, B.G. Comparative Politics: Theory and Method, Basingstoke, Palgrave, 1998 at 14. and to do so in their natural environment. One of the challenges for legal researchers, less so for criminological or sociological researchers who are often trained more fully in this regard, is the need to engage with theory before moving on to the next stage of research design. By theory I mean the explanations that have been posited in the academic literature for why or how something is the way that it is, or claims that suggest relationships between certain things.³²x But interestingly, stance, or more accurately epistemology is of less significance to this research method than to many others. Case study method links the research question, research design, analysis and logic of inference to such an extent that is can accommodate a range of epistemological traditions from the realist to relativist/interpretivist. And thus scholars who consider that there are facts independent of our interpretation of them (in essence, hard facts operating in an objective reality) and scholars who consider all ‘facts’ to be local interpretations constructed through our own lenses, are able to operate within a case study framework. Having said that their choice of data sources, and their approach to data generation and analysis may well vary considerably. In this context a theory is a relatively precise speculative answer to a research question, which may have been developed by undertaking a study or by analysing others’ studies (a meta-analysis). And theories can be converted into hypotheses when considered in the light of a new research question. The use of theory is exemplified by Uriksen and Dadalauri’s case study on tax policy reform in Georgia which aimed to answer why and how Georgia initiated and managed to implement quite radical and substantial tax reforms between 1991 and 2005 and in doing so sought to interrogate theoretical explanations about the nature of policy reform in developing countries and further to develop a model that could be tested in other post-Soviet states.³³x Ulriksen and Dadalauri, ibid. See further Dadalauri, N. Tax Policy Formation and the Transnationalizationof the Public Policy Arena; A Case Study of Georgia, Aarhus, Politica, 2011) Georgia was selected as a crucial case. The reasoning for this and also for the methods employed in this study are elegantly set out in the article cited above. In our hypothetical case study it may be possible, for example, to test the theory that mediated agreements lead to less acrimonious relationships between the divorced couple than do lawyer negotiated agreements. One could examine the theory that family mediation is a cheaper and faster alternative to lawyer negotiated settlements and that those mediated agreements are more durable and better suited to family circumstances. In doing so, one may test existing theories and/or to develop a new theory. Or one could test in the chosen context a single theory that is dominant or particularly novel.
  Our researcher could undertake further reading of the literature to add to these hypotheses and to refine them and eliminate those that are no longer plausible in the light of more detailed investigation. This is known as setting out ‘priors’, prior explanations raised in the academic literature.³⁴x On the importance of the identification of priors see Beach, D. and Pedersen, R.B., ‘What is process tracing actually tracing? The three variants of process tracing methods and their uses and limitations’ Sept 1-4 2011 The American Political Science Association Annual Meeting, Seattle, WA as cited by Ulriksen and Dadalauri, ibid; and further Gerring, 2007 ibid. The researcher’s next task would be to consider what she would expect to observe in the study, were any of these hypotheses true (the observable implications of the hypotheses), for example, our researcher would expect to see that divorces conducted using family mediation would be settled through mediation, that the settlement process would cost less and be concluded quicker than in lawyer-led divorces, and that those that used family mediation would be more able to engage in constructive dialogue post-divorce and to renegotiate arrangements in respect of children without the need to resort to lawyers or to the courts. The researcher may also draw up a hypothesis that the agreements would be more durable and the outcomes for children more positive. This pre-emptive delineation of as many possible observable implications, and how they could be measured, would allow the researcher to plan how to conduct the study and to adopt an appropriate design more likely to lead to a robust answer to the research question.³⁵x For more assistance on extracting observable implications and considering their measurement see Epstein and King, 2002, 70-76. But one needs to be able to articulate the theory converted into hypotheses with clarity, in order for an observation protocol to be developed. Further, it helps the research design if the researcher is able to pose rival theories or explanations so as to design the study to test for plausible alternative explanations too. For example, one of the rival theories in the family mediation study is that the kind of people who use mediation through to conclusion and the kind of people who either refuse to do so or who drop out without reaching a settlement are different, and those drawn to mediation are more consensus driven and better able to communicate with their spouse than are those who do not. This rival hypothesis would alert the researcher to the need to design the study to examine those who do conclude mediated agreements and those who do not in order to analyse this rival claim. A case study allows for the examination of complex interrelationships between variables in situ, and the theory helps to identify what those variables are.

  C. The Selection of the ‘Case’ or ‘Cases’

  Definitionally this is when it gets somewhat complicated, as ‘case’ can easily become confused with a ‘legal case’ and further a ‘case‘ can sometimes be confused with the same word used in a different context in quantitative research, a case meaning a single observation or a single data point. This has led some, such as Gerring, to suggest that it may be more accurate to refer to a case study as a unit study so as to underline that this type of study examines multiple things within one unit rather than examining one data source in one context.³⁶x See Gerring, 2004, 342. See further King, Keohane, and Verba, 1994, 76-77. The selection of the case or cases is a profound one in any study of this kind. A case may be selected because it is critical to the research question, it is typical, atypical, it provides a longitudinal opportunity (study over time), or it is revelatory meaning that it allows insight where previously this has not be possible.³⁷x See Yin, 2014, 51. The nature of the case, its boundaries and features and why it was selected should be set out clearly.³⁸x Gerring, 2004, 344. There are often difficulties in establishing the boundaries of the case, the phenomenon under study, and the context that provides a background to the phenomenon but is not itself the object of enquiry. Gerring articulates this as the formal case (the phenomenon) and the informal cases (the penumbra of phenomena which are the context but which will need to be explored in a less formal way so as to distinguish the boundaries of the formal case). The informal units are peripheral, but may have bearing on the formal unit or case, and by considering these informal units at the beginning of the study, and close to its conclusion it will help the researcher to work out what is particular about the unit, and what is transferable to other units. The selection of the case should be guided by the extent to which this location in space, focus and/or time lends itself to construct validity, internal validity, external validity and reliability of design in respect to the research question.³⁹x See Yin, 2014, chapter 2.
  Case studies are particularly prone to selection bias, meaning that the case is selected on the basis of the dependent variable rather than on the basis of the independent variable – selected because of an effect that has been noted rather than its cause when the nature of the cause is the real object of many ‘how’ or ‘why’ research study questions.⁴⁰x See Geddes, B. Paradigms and Sand Castle: Theory Building and Research Design in Comparative Politics, Ann Arbor, University of Michigan Press, 2005. For example, in a legal context if we wanted to examine rival explanations for how a particular legislative reform, for example the introduction of same sex marriage in England and Wales, has had an impact on community cohesion between different religious and community groups, it may seem, on the face of it, a good idea to select a town like Brighton with a vibrant LGBT community as the case to be studied. After-all, the uptake of same-sex marriage has been very high in Brighton and so it could be considered to be a key site of study. However, Brighton is well known as a LGBT friendly town and people drawn to live there would tend to be very positive about the introduction of same-sex marriage. If the reason for the study was to consider whether tensions have emerged between community and religious groups with different views on marriage, then Brighton would not likely give much opportunity to examine these issues. It was LGBT friendly before the reform and it continues to be. And community and religious groups have worked well together before and after the change in the law. By selecting the case on the basis of the effect of the changed legal landscape, the high numbers of gay and lesbian marriages, we may have selected a case that is atypical or simply a poor unit within which to view the causes or the influences that led to the legislative reform. Researchers are prone to make this mistake when undertaking a deductive study to test the congruence of rival hypotheses in a context where they have insufficient knowledge about the independent variable (the causes) that gave rise to, say, the change in the law. There may be other factors to consider too: in our family mediation study, the researcher may choose to steer away from London as the case study location, if she is interested in ‘typical’ divorces given that London has a much greater than average number of high net worth divorces that include very large sums of money and property portfolios, in addition to many divorces involving non-British couples who married abroad. A solid grasp of the literature can help to alleviate this possibility of incorrect case selection along with detailed consideration of the relevant features of a range of possible case studies prior to final selection.
  Does selecting a multiple-case study limit the likelihood of such problems, further should the study be at one point in time or a repeated measure at different periods of time? A case study may be designed so as to allow for cross sectional analysis between two or more cases, and further a temporal variation may be introduced into this form of analysis too. Our researcher could study family mediation over time: the same divorces pre, during and post settlement and then later again to examine durability. Single, multiple or cross-sectional case studies often serve different purposes. Multiple case studies are more likely to be used when the causal relationship between an independent and dependent variables are being analysed, so that the interaction of the variables in different environments can be examined comparatively in different contexts (in an experimental protocol one would be able to manipulate the conditions so as to test the variables and thus the different hypotheses). For example, if the study was examining the relationship between violent crime rates and criminal justice sentencing policy to examine whether tougher criminal penalties for violent crime lead to a reduction in violent crime rates, and if tougher criminal penalties for lesser offences led to greater imprisonment levels and greater recidivism including an escalation of violent crime, then a multi-jurisdictional case study may allow for a better assessment of those by permitting different combinations of variables to be compared as against each other. In our family mediation study, the researcher may choose to use London as a crucial case study (with its unusual profile of divorcing spouses with a very wide range of asset values) alongside a more typical rural and a more typical urban location to consider the hypotheses under different conditions. However, it may also be possible to test the hypotheses in a single case study by charting the relationship between the variables over time, with particular attention being paid to the points in time when sentencing policy changed or crime rates dropped or raised, or when family mediation was first introduced, when it became established as a compulsory part of the system.⁴¹x But the difficulty with multi case studies is that specified conditions or features within individual cases may have more influence on the variables being studied than the variables that one are analysing across the studies. This may lead one to draw erroneous conclusions about causality. This single case study also illustrates the independent variable problem: to what extent are violent crime rates and/or sentencing policy more likely attributable to other societal changes evident at different points in time than each other? Without knowledge of this it is difficult to proceed.
  Sometimes the extent to which a case study is referred to as single or multiple is a matter of nomenclature, for example, Elliott and Kling’s study⁴²x Elliott, M. and Kling, R. ‘Organizational Usability of Digital Libraries: Case Study of Legal Research in Civil and Criminal Courts’ (1997) Vol 48 (11) Journal of the American Society for Information Science 1023-1035. on the organisational usability of digital libraries, a case study of legal research in civil and criminal courts, could be described as a single case study (as in their study) because it addresses digital libraries in one context – legal research in courts – it is also geographically bounded to the Los Angeles County, but data is collected from a number of courts and thus it could be argued to be a multiple case study if each court were considered to be a case. The important distinction, however, is how that data are treated: if the data are pooled and analysed as a single unit then the case study is generally considered to be a single unit or single case study, if the data are analysed comparatively as between the sites of collection then it would generally be considered to be a multiple unit or multiple case study. Where comparisons are being made over time but within a unit then the terminology is often that of a single unit as data is both compared and pooled too.
  Single case study research is considered to be an excellent vehicle for exploratory and developmental research (as evidenced by the Georgia tax policy study and Dnes analysis of the nature of a particular type of contract – franchise contracts in the UK⁴³x See Dnes, A.W. ‘A Case-Study Analysis of Franchise Contracts’ (1993) Vol. 28 Journal of Legal Studies 367-393. or Latour’s Conseil d’Etat study mentioned above⁴⁴x See Latour, B. (2010) above. ), confirmatory research necessitates a design that allows the researcher robustly to test a small number of hypotheses forensically and it may be advisable to consider a multiple case study method to achieve this aim.⁴⁵x Although note that Gerring, 2004, at 347 indicates that a single case study may credibly make causal claims, if, for example, the case has been selected as it is particularly representative of others or it is a critical or crucial case, see further: Eckstein, H. (1975) ‘Case Studies and Theory in Political Science’ in Regarding Politics: Essays on Political Theory, Stability, and Change, Berkeley: University of California Press, 1992). The family mediation case study conducted in London and a more typical urban and rural area is a good example of this, given that the types of divorce cases are likely to be quite different, the context is also different too, and so cost, duration, durability and acrimony could be tested under different conditions to see whether they held true in all conditions or were context or divorce type dependent. Sometimes researchers are inclined to use a multiple case study approach in the expectation that more cases (units) will provide more data and more comparable data that can be used to derive robust findings. However, data collected across multiple case studies is less rather than more likely to be comparable as the conditions within the case study cannot be manipulated or controlled by the researcher and yet environment is expected to have an impact on the data.⁴⁶x Gerring indicates that researchers tend to face the choice between knowing more about less or less about more, 2014, 348. Where more than one case is selected, each subsequent addition should provide a more complete and accurate picture in respect of the research question, instead of attempting to provide greater representativeness (as indicated above, this is not the purpose of case study method).⁴⁷x See Small, 2009, 24-26. If one wishes to make comparisons between case studies it is important to adhere closely to comparative methodology in the definition of the cases to be selected, the analysis of the relevant similarities and differences between those cases, the data to be collected, compared and why, and the likely limitations of the cross unit comparison.⁴⁸x See Gerring, 2004 at 348. For guidance on comparative methodology in a legal context see: Van Hoecke, M. (2004) Epistemology and Methodology of Comparative Law, Portland, Oregon: Hart Publishing. Multiple cases studies are more difficult to accomplish successfully, and it is advisable to work out clearly what each of the case studies will contribute to answering the research question before finalising those to be the subject of the enquiry. The most appropriate design will be dependent on the research question selected and the hypotheses or propositions under investigation through the case study method.⁴⁹x For a detailed discussion see Gerring, 2004 at 343.

  D. The Selection of Data Sources, Data Generation and Collection

  Strong research design logically links the research question(s) with the hypotheses, with the data generation and collection methods, which in turn should be logically linked with the data analysis methods employed too. By now the researcher is likely to have a very good idea of the types of data that may be relevant to the study (derived from documents, people, extant statistics, other artefacts like images),⁵⁰x Yin sets out six sources of evidence: documents, archival records, interviews, direct observations, participant observation, physical artefacts, Yin, 2014, 105-118, and four principles of data collection: multiple sources of data; creation of a case study database; maintain the chain of evidence; exercise care when using data from electronic sources, at 118-129. having identified the substance, form and purpose of the question, the rival hypotheses that may contribute to answering the question, and the observable implications of those hypotheses and how they may be measured. Our researcher having drawn her broad question as ‘how have the reforms to family mediation affected the way in which divorces are conducted?’ and narrowed down the case study to a geographic location(s) or a type of divorcing couples or divorce context, will have considered the possible data sources as including the spouses, their children, the family mediators and lawyers, court files, mediation and lawyer negotiated agreements, official statistics and more. And so it should be possible to chart how the data collection and data analysis methods all fit together so as to allow the observable implications to be explored, the hypotheses proved, amended, or disproved and a rounded, reliable answer to the research question be achieved.
  But this design phase may also go beyond the identification and selection of data sources, it requires choices to be made about how data will be collected and measured.⁵¹x See Yin, 2014, chapter 3 for more information on what one needs to do before data collection begins, and chapter 4 on data collection itself. Each choice that is made, consciously, or unconsciously, will have an effect on the data that is captured and the reliability and validity of that data. And this in turn will affect the outcome of the study. Consequently, the design should be scrutinised to uncover the biases that may become entrenched within it, the study redesigned where necessary to eliminate or limit bias and any remaining biases be taken into account during the later analysis and reporting phases.⁵²x See Small, 2009, 12-15. This will require a degree of reflection on data type (strengths and weaknesses in allowing observable implications to be explored), data selection (all data, if not all then what process is being used to select it and how may that skew the findings, known as selection bias); data collection (how is the data being derived, and is it raw data or is it material that requires a judgement to be made, for example how will we measure ‘satisfaction or ‘acrimony’ or ‘durability of an agreement’, how reliable and valid is the data collection instrument⁵³x For a discussion of reliability and validity in measurement see Webley, 2010 and further Epstein and King, 2002, 80-99. ) and later too data analysis. It is also worth piloting each data collection exercise with a small number of observations so as to allow defects to be worked out, and experience in the field to allow for redesign too. And then one should collect as much data on each of the possible observable implications as is practicable, including data of different types generated or collected via different methods so as to allow for triangulation in respect of each hypothesis. For example, in Elliott and King’s study they collected data via observation, participant observation and interviews, analysed court documents and legal technology documents;⁵⁴x Elliott and Kling, ibid at 1025. in Dnes’ study the data included franchise agreements and contracts, financial accounts and other financial data, interviews;⁵⁵x Dnes, ibid at 369-370. and in Dadalauri’s study data were derived from primary sources (policy proposals, experts’ recommendations and the minutes of parliamentary sessions) and secondary sources (reports, media briefs, statistical sources) plus from semi-structured interviews with key actors in the policy process.⁵⁶x Ulrikesn and Dadalauri, ibid at 13. Case study method necessitates a measure of flexibility in research design to allow for new knowledge to shape and improve the starting design, but that does not reduce the need for a robust design plan at the inception of the study. The design needs to be scrupulously documented, including challenges faced and amendments made so as to aid others to analyse the validity of the research design and to assess the extent to which the study findings are reliable and robust.⁵⁷x For a discussion see King, Keohane and Verba, 1994, 12. Epstein and King suggest that legal scholars give the same attention to the recording, storage and analysis of data as they would expect of the police and prosecutors when securing the chain of evidence in a criminal case.⁵⁸x Epstein and King, 2002, 24. And it is to data analysis that we now turn.

• 3. Case Study Method: Use of the Data, Inferences and Finding Meaning

  Case study findings are reached through a process of logical valid inferences regardless of whether the data collected and analysed is qualitative, quantitative or both.⁵⁹x For a discussion on this point see King, Keohane and Verba, 1994, chapter 1 ‘The Science in Social Science’. But first the data must usually be described in summary form, before being subjected to further analysis to consider what the data indicates about the various hypotheses and their observable implications in this case context. Subsequently, it is possible to attempt to derive descriptive inferences that suggest what these data on observable instances indicate about non-observables ones, in other words what findings one considers to be transferable to a non-observed context. The analysis may also allow for causal inferences to be made, that explain what effects would be expected to occur if certain conditions were fulfilled in this or another context. This is not dissimilar to data analysis in other types of empirical legal research and therefore it is considered only briefly here. However, case study method is structured with triangulation of data at the fore, allowing the researcher to reach robust findings reached by integrating analysis from multiple data points gathered using different methods. This section will briefly address data analysis, the drawing of inferences and the importance of demonstrating one’s working out, in turn.

  A. Data Analysis

  The first stage of data analysis is often validly to summarise the data collected in the light of the research question and hypotheses and anticipated observable implications, to summarise the numbers (mean, median, mode, standard deviation, range) and to summarise the text (for example, categorise and consider relationships between categories, or code and consider the frequency or codes).⁶⁰x See Epstein and King, 2002, 25-29 for more information on quantitative data description. Different types of data will often be analysed using different methods or traditions, as illustrated by the way in which legal cases are analysed according to traditions accepted by lawyers, which is distinct from legal analysis of legislation, and policy analysis of policy documents: survey data would be analysed statistically, text based data (interviews, documents etc.) via the mode of analysis selected to interrogate and derive meaning from language, for example via grounded theory method, thematic coding, content analysis, hermeneutics etc.⁶¹x For a discussion of the different methods of text based data analysis see Yin, 2014, chapter 5 or Webley, 2010. There are a range of general strategies open to the researchers, some of which focus on the theoretical propositions, others aim to develop thick description, others still examine plausible rival explanations.⁶²x See Yin, 2014, 136-142. Findings are considered robust where they are evidenced via multiple stands of data and its analysis. The use of multiple data sources to test each hypothesis allows the researcher to build up a thoroughly nuanced picture of the extent to which each hypothesis is sustained, needs to be refined, or rejected. The analysis will be conducted in the light of the research question parameters and also the hypotheses being examined by the research, as exemplified by the discussion in the Georgia tax policy case study.⁶³x Ulriksen and Dadalauri, ibid. This process is likely to be iterative, in that data will often be analysed as one phase of data collection is complete and any lessons learned from that may lead to some reframing of the research question, reconsideration of the hypotheses, and amendments to the next phase of data collection yet to begin. The key is that, as with all social science methods, amendments to the question and methods, the analysis of the data and the inferences drawn from the data should be publicly explained and in sufficient detail so that they are replicable by others on the basis of the information provided in the write-up of the study; King, Keohane and Verba remind us that inferences lead to uncertain conclusions – inferences are not proven facts, they are propositions being advanced that are available to be tested by others.⁶⁴x King, Keohane and Verba, 1994, 8. Conclusions remain tentative until replicated validly and consistently. The science and the rules of inference are important in allowing us to judge the validity and reliability of the findings, and these are closely interwoven with the research design and execution of the study.

  B. Inferences

  Case studies are often considered to be more useful when seeking to derive descriptive rather than causal inferences, as the researcher is not able to manipulate the environment so as to test propositions in such a way as to be sure that causal relationships have satisfactorily been established. Descriptive inferences are ‘the process of using the facts we know to learn about facts we do not know’, by describing something that has been observed and inferring under what circumstances a similar pattern or occurrence may occur in a carefully defined unobserved situation.⁶⁵x Epstein and King, 2002, 29. For example, if in our hypothetical family mediation study we learned that greater numbers of the divorcing clients who we interviewed/observed before the introduction of the compulsory mediation information and assessment meetings were aware that there was state funding available for family mediation, compared with the divorcing clients who we interviewed/observed after the introduction of these meetings then we may infer that this finding was likely to apply to divorcing clients outside our observed group too (all other things being equal). We do not know for certain that is accurate, as we only have data from our study participants, but our description of our findings has led us to infer something about those outside our observation group. Many doctoral candidates and early career academics baulk at the suggestion that descriptive inference is a valuable mode of analysis, as they associate ‘descriptive’ with the less positive feedback that they may have received in earlier work. But the pursuit of descriptive inferences is not a low-level aspiration in a context in which little is known about the case under scrutiny. Descriptive inferences allow for categorisation of findings which may lead on to further theory building and theory testing, categorisation goes to the heart of analysis development. So our finding above begs the question ‘why is this so?’ and we could either extend our study to answer this sub-question, or leave that for a later study.
  In some instances the inferences a researcher wishes to draw may be causal ones that infer an effect that will be caused by a set of defined factors occurring together. As an example, in our family mediation study we may wish to examine whether family mediation is more likely to be successful for couples with relatively similar educational backgrounds, medium to high incomes with both spouses in full-time employment, when compared with those who have unequal educational backgrounds, incomes and job-statutes and with low incomes. Where causal inferences are the point of the study, it may be possible to develop these with a well-chosen cross-case multi-case case study design. However, a causal inference first requires the identification of a causal mechanism (the process by which dependent variable A is affected by independent variable B, for example the causal mechanism for a defendant in the UK to be released from pre-trial detention (variable A) is a bail hearing in court (variable B)).⁶⁶x See Epstein and King, 2002, 34-37 for help distinguishing between causal mechanisms and variables and causal effects. Case studies are often a really good means by which these mechanisms, or processes, may be uncovered – known as ‘process tracing’ whereby the researcher charts in detail the relationships between two or more variables and explores these connections to deduce those that are causal and those more likely to be coincidental.⁶⁷x See Gerring, 2004, 348 and further Roberts, C., The Logic of Historical Explanation, Pennsylvania State University Press 1996, 66. Further, a single case study may allow a researcher to interrogate extant explanations that suggest causal implications, in other words to test predictions about what will happen in particular situations (assuming those situations are observable as part of the case study). This is known as ‘pattern-matching’.⁶⁸x See Gerring, 2004, 348 and further Campbell, D. T. [1975] “‘Degrees of Freedom’ and the Case Study” in E. Samuel Overnman (ed) Methodology and Epistemology in Social Science, Chicago, University of Chicago Press: 1988, 380. This is where clarity about the purpose of the study becomes particularly important, as certain conditions will need to have been built into the research design for some analytical techniques.⁶⁹x Yin, 2014, 142-168, provides five different analytical techniques: pattern-matching; explanation building; time-series analysis; logic models; and cross-case synthesis and suggests that after this phase the researcher will likely move on to work through all plausible alternative conclusions to examine whether the most likely conclusion is the only conclusion. The study will need to be designed with a very clear and narrow focus to achieve its aims.
  As indicated above, case studies may be entirely self-contained studies that provide in-depth knowledge of a single unit of analysis, but more often than not the researcher will wish for those findings to be considered applicable to situations that she/he has not observed. The challenge is to explain which findings are particular to the case study and which elements of the findings are relevant beyond the case study’s boundaries.⁷⁰x Gerring, 2004, 345. It may be difficult to define this with precision, but where there is ambiguity it is safer to over explain and to over report the ambiguity and the possible range of inferences and their limitations rather than to over simplify and obfuscate the difficulty in reaching definitive findings.⁷¹x See Gerring, 2004, 346. Legal researchers are sometimes criticised for being vague in their explanations of the target of their inferences (to which other unobserved situations do these findings apply, and why?), or worse still their claims in the absence of evidence to prove that their inference is generalizable to a wide variety of situations.⁷²x Epstein and King, 2002, 31. This may be a function of lawyers’ professional training as advocates, who in presentation would seek to persuade others to accept their position and who would gloss over inconvenient precedents. But lawyers are also trained to be forensic in seeking out the weaknesses of their arguments as well as those of their opponents and by harnessing these skills in the presentation of their case study findings; they should be able to display the highest standards of scientific reporting. Some of the ambiguities associated with inferences may be avoided if, as Gerring suggests, the scholar specifies clearly which propositions apply to which novel circumstances and exhibit and explain the evidence upon which this contention is based.⁷³x Gerring, J. (2001) Social Science Methodology: A Criterial Framework, Cambridge, Cambridge University Press, 90-99. In other words, do the findings relate only to this case, are they intended to relate more broadly to similar cases and if so what marks out other situations as similar? Is similarity about time frame, location, a certain set of markers such types of participants, socio-economic, legal or political factors? And what is one’s evidence in support of this? The burden of proof always rests with the researcher. We shall turn to this in the next sub-section.

  C. Reporting Findings

  It can be challenging to know how to report one’s findings in an article or thesis, which is unsurprising when one considers that little attention is paid to this aspect of scholarship on doctoral legal programmes in many jurisdictions.⁷⁴x See further Epstein and King, 2002. For assistance in how to report on cases studies and writing up and presentational considerations, including audience and purpose, see Yin, 2014, chapter 6. The rule is that one must provide as much detail as possible, at least enough to allow someone else to be able to replicate the study using only the information provided. Further, there needs to be sufficient discussion of the decisions taken, challenges faced and the consequent limitations of the findings so as to allow others to evaluate the reliability of one’s findings. As King et al. note: report uncertainty, be sceptical about causality, and consider rival hypotheses.⁷⁵x King et al, 1996, 30-33. The process of interrogating one’s own decisions and inferences and reporting on them in full in the article or thesis may allow one to avoid the invidious charge made of many other legal scholars’ work.⁷⁶x Epstein and King, 2002, 6-7. One suggestion is that legal scholars may wish to look for the weakest link in their chain of reasoning, something which lawyers are trained to do in a legal context, and then estimate how certain they are of their findings taking that weakness into account.⁷⁷x Epstein and King, 2002, 50. However, other aspects of our professional training sometimes come into conflict with this approach: a research study is not an act of advocacy, and training as a lawyer may derail the process of empirical enquiry when lawyers unconsciously act for the client in their head and seek to persuade the outside world that their client’s view is a valid one, rather than to act as a legal social-scientist and demonstrate to other social scientists the extent to which their findings are valid, robust, reliable, and subject to limitation.⁷⁸x See further Miller, A. S. ‘The Myth of Objectivity in Legal Research and Writing’ (1969) Vol. 18 Catholic University Law Review 290. For example, a researcher who is more in favour or less in favour of family mediation may inadvertently confirm their stance and steps need to be taken to lessen this risk. This role conflict is particularly problematic given that empirical legal research may lead to legal reform affecting large sections of the population and the findings confidently exhorted in the literature may be used to justify policy changes.⁷⁹x Epstein and King, 2002, 8-10. Further, even if the research were not to be read outside of an academic environment, it is incumbent on all academics to produce research that is reliable and robust, lawyers are quick to critique legal scholarship that had been poorly executed and socio-legal scholarship should be treated no differently.

• 4. Conclusions: Why (Not to) Use Case Study Method?

  Case study method is a powerful and engaging approach to research that has real utility in socio-legal and criminological research even if it has to-date been relatively little used. Our reticence to use it may be explained by the need for a researcher to be sufficiently adept with a range of social science research methods; (non-legal) empirical methods have historically had little treatment within undergraduate legal courses and relatively little attention even at a postgraduate level.⁸⁰x Genn, H., Partington, M, and Wheeler, S. (2006) Law in the Real World: Improving Our Understanding of How Law Works Final Report and Recommendations, London: The Nuffield Foundation. Further, doctoral supervisors may feel inadequate to the task of supervising doctoral students proposing to undertake research through case study method and steer them towards a more standard mixed method approach such as a survey coupled with some interviews, or away from non-legal empirical methods altogether. But with some training, and a high degree of planning it is perfectly possible to undertake a good quality case study in a legal context and we can learn much from them. They are also an ideal means to focus on the particular and yet to draw analytical inferences to similar contexts too, something which lawyers are trained to do throughout their studies and a skill which they can bring to bear on a broader range of data than they otherwise often do.
  However, case study method is far more than focusing on a single situation, or ‘case’, it is far more than providing a temporal or physical boundary to our research endeavour. It requires us to adopt a structured and reflective approach to research design in many instances, to consider pre-emptively possible explanations (hypotheses) and rival propositions and to engage with theory at an early stage in a study. In a legal context case studies are generally, if not exclusively, more effective when: seeking to make descriptive rather than causal inferences; examining issues in depth rather than broadly and when the researcher is seeking to examine multiple sources of data so as to make comparisons within a case rather than between multiple cases. Further, they are also often more effective for seeking causal mechanisms rather than causal effects; for research that is exploratory rather than confirmatory; and when variations within the case selected are important for the study of the phenomenon.⁸¹x Gerring, 2004, 352. They are extremely useful when analysing how those involved in law and policy-making, the application of legal rules and procedures perceive these processes, how they react to them and how this influences the effectiveness of those rules, processes and procedures. The research process is an iterative and creative one that engages lawyers’ considerable analytical skills. As such case study method is worthy of a larger presence within the legal academic empirical tool-kit.

Noten

• 1 For insight into the extent to which legal researchers undertake empirical research and the lack of clarity around empirical methods in law see Epstein, L. and King, G. (2002) ‘The Rules of Inference’ Vol. 69 No. 1 The University of Chicago Law Review 1-133 at 3-6, Part I.

• 2 Yin, R. K. (2014) Case Study Research Design and Methods (5th edn.) Sage Publications, 12-14.

• 3 See Hamel, J. with Dufour, S. and Fortin, D. (1993) Qualitative Research Methods Volume 2, Sage Publications, ch 1.

• 4 See Gerring, J. Case Study Research: Principles and Practices, Cambridge, Cambridge University Press, 2007 at 6. For a further discussion see Gerring, J. ‘What is a case study and what is it good for?’ (2004) Vol. 98 No 2 American Political Science Review 341-354. This looseness in definition in a legal context may perhaps be linked to confusion as between teaching and research case studies; some traditions in legal education employ a teaching method known also as ‘case study method’ which operates quite differently from its research counterpart. For a discussion of the differences between teaching and research case studies see Yin, 2014, 20 and for a discussion of teaching case studies see Ellet, W., (2007) The Case Study Handbook: How to Read, Discuss, and Write Persuasively About Cases, Boston MA, Harvard Business Review Press; Garvin, D. A. (2003) ‘Making the Case: Professional Education for the World of Practice.’ (Sept–Oct) Harvard Magazine 56-65.

• 5 ‘n’ (number) is used to denote the number of observations in the study, N is used to describe the total number within the population when n denotes the sample observed.

• 6 For a discussion of the state of empirical research in law see Epstein and King, 2002.

• 7 See Webley, L. (2010) ‘Part III Doing Empirical Legal Studies Research Chapter 38 - Qualitative Approaches to ELS’ in Cane, P. and Kritzer, H. (eds.) Oxford Handbook of Empirical Legal Studies, Oxford, Oxford University Press.

• 8 For a discussion about the differences between numerical (quantitative) and non-numerical (qualitative) data see Webley, id; Epstein and King, 2002, at 2-3; King,G., Keohane, R. O., and Verba, S., Designing Social Inquiry: Scientific Inference in Qualitative Research, Princeton NJ, Princeton University Press, 1994 at 6.

• 9 Mascini, P., van Oorschot, I., Weenink, D. and Schippers, G., (2016) ‘Understanding judges’ choices of sentence types as interpretative work: An explorative study in a Dutch police court’, (37) (1) Recht der Werkelijkheid 32-49.

• 10 Yin, 2014: xxxi and further 16-17.

• 11 Triangulation is the term used to explain that a research question is considered from as many different standpoints as possible, using as many different data types as possible to permit a holistic examination of the question to see which explanations, if any, remain consistent across all data sources.

• 12 Yin, 2014: 5-6.

• 13 See Dul, J. and Hak, T. (2008) Case Study Methodology in Business Research, Oxford: Elsevier 8-11, 30-59.

• 14 Latour, B. (2010) The Making of Law: An Ethnography of the Conseil D’Etat, Cambridge: Polity Press.

• 15 For a discussion of ethnomethodological aims to study practical life as experienced in context as an end in itself, as experience is subjective and situational, see Small, M.L. ‘‘How many cases do I need?’ On science and the logic of case selection in field-based research’ (2009) Vol. 10 (1) Ethnography 5, 18.

• 16 For greater insight on this point see Lipset, S. M., Trow, M. and Coleman, J.S. (1956) Union Democracy: The Internal Politics of the International Typographical Union, New York: New York Free Press at 419-420; Yin, 2014, 21. For a discussion of the problems inherent in aping quantitative terminology in qualitative work see, Small, 2009, 10, and at 19 for further reading on the logic of case study selection and further reading on extended case method.

• 17 Gerring, 2007, 346.

• 18 Although note that there are some scholars who believe that case study method can include elements of experimental testing, for example, Gerring, J. and McDermott, R. (2007) ‘An Experimental Template for Case Study Research’ Vol. 51 No. 3 American Journal of Political Science 688-701. One such study in law that has been described by some, if not by the researchers themselves, as a case study did include an experimental design within the battery of methods employed see: Moorhead, R., Sherr, A., Webley, L., Rogers, S., Sherr, L., Paterson, A. & Domberger, S. (2001) Quality and Cost: Final Report on the Contracting of Civil Non-Family Advice and Assistance Pilot (Norwich: The Stationery Office).

• 19 Campbell, D.T. Foreword in Yin, 2014 xviii.

• 20 Seawright, J. and Gerring, J. (2008) ‘Case Selection Techniques in Case Study Research: A Menu of Qualitative and Quantitative Options’ Vol. 61 No. 2 Political Research Quarterly 294-308.

• 21 Yin, 2014: chapter 1.

• 22 For a discussion see: Ryle, G. (1949). The Concept of Mind. London: Hutchinson; Lincoln, Y.S. and Guba, E.G. (1985). Naturalistic Inquiry. Newbury Park, CA: Sage Publications; Holloway, I. (1997). Basic Concepts for Qualitative Research. London: Blackwell Science.

• 23 Confirmation or interpretive bias of data is something we all have grapple with, as the natural human tendency is to place more weight on evidence that confirms our view than on evidence that contradicts it. Strong research design can assist with counter-balancing this to some extent, including the transformation of any expected finding into a hypothesis that one then seeks to falsify rather than to confirm.

• 24 See Yin, 2014, chapter 1 for more information on the role of the literature review.

• 25 For more assistance with legal research questions refer to Epstein, L. (1995) Studying Law and the Courts in Lee Epstein (ed) Contemplating Courts, Cong Q, 1, 3-5.

• 26 See King, Keohane and Verba, 1994, 15.

• 27 See Yin, 2014, at 11, and see further Campbell, J.P., Daft, R.L., Hulin, C. L. (1982) What to Study: Generating and Developing Research Questions (Studying Organizations), Sage Publications, for further thoughts on research questions

• 28 On process tracing see: Collier, D. ‘Understanding Process Tracing’ Vol. 44 Political Science and Politics 823-830 and George, A. L. and Bennett, A. Case Studies and Theory Development in the Social Sciences, Cambridge, MA, MIT Press, 2005. For an excellent insight into how this has been used in a legal and policy context with reference to changes in Georgia’s tax laws see Ulriksen, M.S, and Dadalauri, N. ‘Single Studies and Theory-testing: The Knots and Dots of the Process-tracing Method’ (2014) International Journal of Social Research Methodology 1- 17.

• 29 See Blatter, J. and Haverland, M. Designing Case Studies, Basingstoke, Palgrave MacMillan, 2012 at 145 who consider process tracing involves inductive reasoning to build theory and congruence testing involves deductive reasoning to test theories.

• 30 Epstein and King suggest a range of possible purposes, at 59, including: to explore something that has not previously been studied; to attempt to settle a debate that has been ongoing within the literature; to examine a well-considered question but in a new way; to collect and analyse new data to seek to confirm or refute previous findings; to analyse an existing data set in a new or better way to seek to confirm or refute previous findings or to develop new ones.

• 31 Peters, B.G. Comparative Politics: Theory and Method, Basingstoke, Palgrave, 1998 at 14.

• 32 But interestingly, stance, or more accurately epistemology is of less significance to this research method than to many others. Case study method links the research question, research design, analysis and logic of inference to such an extent that is can accommodate a range of epistemological traditions from the realist to relativist/interpretivist. And thus scholars who consider that there are facts independent of our interpretation of them (in essence, hard facts operating in an objective reality) and scholars who consider all ‘facts’ to be local interpretations constructed through our own lenses, are able to operate within a case study framework. Having said that their choice of data sources, and their approach to data generation and analysis may well vary considerably.

• 33 Ulriksen and Dadalauri, ibid. See further Dadalauri, N. Tax Policy Formation and the Transnationalizationof the Public Policy Arena; A Case Study of Georgia, Aarhus, Politica, 2011) Georgia was selected as a crucial case. The reasoning for this and also for the methods employed in this study are elegantly set out in the article cited above.

• 34 On the importance of the identification of priors see Beach, D. and Pedersen, R.B., ‘What is process tracing actually tracing? The three variants of process tracing methods and their uses and limitations’ Sept 1-4 2011 The American Political Science Association Annual Meeting, Seattle, WA as cited by Ulriksen and Dadalauri, ibid; and further Gerring, 2007 ibid.

• 35 For more assistance on extracting observable implications and considering their measurement see Epstein and King, 2002, 70-76.

• 36 See Gerring, 2004, 342. See further King, Keohane, and Verba, 1994, 76-77.

• 37 See Yin, 2014, 51.

• 38 Gerring, 2004, 344. There are often difficulties in establishing the boundaries of the case, the phenomenon under study, and the context that provides a background to the phenomenon but is not itself the object of enquiry. Gerring articulates this as the formal case (the phenomenon) and the informal cases (the penumbra of phenomena which are the context but which will need to be explored in a less formal way so as to distinguish the boundaries of the formal case). The informal units are peripheral, but may have bearing on the formal unit or case, and by considering these informal units at the beginning of the study, and close to its conclusion it will help the researcher to work out what is particular about the unit, and what is transferable to other units.

• 39 See Yin, 2014, chapter 2.

• 40 See Geddes, B. Paradigms and Sand Castle: Theory Building and Research Design in Comparative Politics, Ann Arbor, University of Michigan Press, 2005.

• 41 But the difficulty with multi case studies is that specified conditions or features within individual cases may have more influence on the variables being studied than the variables that one are analysing across the studies. This may lead one to draw erroneous conclusions about causality. This single case study also illustrates the independent variable problem: to what extent are violent crime rates and/or sentencing policy more likely attributable to other societal changes evident at different points in time than each other? Without knowledge of this it is difficult to proceed.

• 42 Elliott, M. and Kling, R. ‘Organizational Usability of Digital Libraries: Case Study of Legal Research in Civil and Criminal Courts’ (1997) Vol 48 (11) Journal of the American Society for Information Science 1023-1035.

• 43 See Dnes, A.W. ‘A Case-Study Analysis of Franchise Contracts’ (1993) Vol. 28 Journal of Legal Studies 367-393.

• 44 See Latour, B. (2010) above.

• 45 Although note that Gerring, 2004, at 347 indicates that a single case study may credibly make causal claims, if, for example, the case has been selected as it is particularly representative of others or it is a critical or crucial case, see further: Eckstein, H. (1975) ‘Case Studies and Theory in Political Science’ in Regarding Politics: Essays on Political Theory, Stability, and Change, Berkeley: University of California Press, 1992).

• 46 Gerring indicates that researchers tend to face the choice between knowing more about less or less about more, 2014, 348.

• 47 See Small, 2009, 24-26.

• 48 See Gerring, 2004 at 348. For guidance on comparative methodology in a legal context see: Van Hoecke, M. (2004) Epistemology and Methodology of Comparative Law, Portland, Oregon: Hart Publishing.

• 49 For a detailed discussion see Gerring, 2004 at 343.

• 50 Yin sets out six sources of evidence: documents, archival records, interviews, direct observations, participant observation, physical artefacts, Yin, 2014, 105-118, and four principles of data collection: multiple sources of data; creation of a case study database; maintain the chain of evidence; exercise care when using data from electronic sources, at 118-129.

• 51 See Yin, 2014, chapter 3 for more information on what one needs to do before data collection begins, and chapter 4 on data collection itself.

• 52 See Small, 2009, 12-15.

• 53 For a discussion of reliability and validity in measurement see Webley, 2010 and further Epstein and King, 2002, 80-99.

• 54 Elliott and Kling, ibid at 1025.

• 55 Dnes, ibid at 369-370.

• 56 Ulrikesn and Dadalauri, ibid at 13.

• 57 For a discussion see King, Keohane and Verba, 1994, 12.

• 58 Epstein and King, 2002, 24.

• 59 For a discussion on this point see King, Keohane and Verba, 1994, chapter 1 ‘The Science in Social Science’.

• 60 See Epstein and King, 2002, 25-29 for more information on quantitative data description.

• 61 For a discussion of the different methods of text based data analysis see Yin, 2014, chapter 5 or Webley, 2010.

• 62 See Yin, 2014, 136-142.

• 63 Ulriksen and Dadalauri, ibid.

• 64 King, Keohane and Verba, 1994, 8.

• 65 Epstein and King, 2002, 29.

• 66 See Epstein and King, 2002, 34-37 for help distinguishing between causal mechanisms and variables and causal effects.

• 67 See Gerring, 2004, 348 and further Roberts, C., The Logic of Historical Explanation, Pennsylvania State University Press 1996, 66.

• 68 See Gerring, 2004, 348 and further Campbell, D. T. [1975] “‘Degrees of Freedom’ and the Case Study” in E. Samuel Overnman (ed) Methodology and Epistemology in Social Science, Chicago, University of Chicago Press: 1988, 380.

• 69 Yin, 2014, 142-168, provides five different analytical techniques: pattern-matching; explanation building; time-series analysis; logic models; and cross-case synthesis and suggests that after this phase the researcher will likely move on to work through all plausible alternative conclusions to examine whether the most likely conclusion is the only conclusion.

• 70 Gerring, 2004, 345.

• 71 See Gerring, 2004, 346.

• 72 Epstein and King, 2002, 31.

• 73 Gerring, J. (2001) Social Science Methodology: A Criterial Framework, Cambridge, Cambridge University Press, 90-99.

• 74 See further Epstein and King, 2002. For assistance in how to report on cases studies and writing up and presentational considerations, including audience and purpose, see Yin, 2014, chapter 6.

• 75 King et al, 1996, 30-33.

• 76 Epstein and King, 2002, 6-7.

• 77 Epstein and King, 2002, 50.

• 78 See further Miller, A. S. ‘The Myth of Objectivity in Legal Research and Writing’ (1969) Vol. 18 Catholic University Law Review 290.

• 79 Epstein and King, 2002, 8-10.

• 80 Genn, H., Partington, M, and Wheeler, S. (2006) Law in the Real World: Improving Our Understanding of How Law Works Final Report and Recommendations, London: The Nuffield Foundation.

• 81 Gerring, 2004, 352.