Designing Social Science Research

A research design is a plan or an approach for carrying out a research project. The plan renders visible the logical structure of the project, the goals of the project, and the strategy to reach these goals. Thus, composing a research design includes two central elements: (1) defining the knowledge goals or purpose of the project and (2) deciding how to proceed to reach the goals. This book presents a number of research designs and shows how they are based on certain combinations of knowledge goals and strategies for data construction. It further presents the process of making a research design, step by step. The goal is to render the readers capable to make an expedient design for their project. Creating a research design comprises more than choosing a method. The choice of method is primarily a question of technique, that is, of what techniques should be used to select, collect, and analyse data on the research phenomenon. The method should not steer the knowledge goals but rather be a tool to achieve those goals. This book will show how different methods can be beneficial in different designs. Still, I do not treat the question of method in any detail. That is a task for the many books discussing research method.

This chapter looks at some of the basic questions concerning the development of scientific knowledge. These are fundamental questions if we wish to carry out a research project. We need to inform ourselves of rules and standards concerning the field of practice that we enter into. I therefore make a short excursion into some basic questions of the theory of science: What are the characteristics of scientific knowledge? Does this knowledge have a different status than our everyday common knowledge? What is the role of cognitive frames in the development of scientific knowledge, and what is the role of theories and scientific concepts? Are theories to be understood as products of data and observations, or do our theoretical lenses shape the way we see the world? Is there such a thing as objective knowledge about society and the world?

Social science is the study of social facts and phenomena. When studying such phenomena, we need an understanding of social acts from within as well as an external explanation of social patterns and processes. In general, we can regard social research as a form of systematic reconstruction of social facts and phenomena. We have established a convincing reconstruction when there is a congruence between on the one hand the ideas and analytical frames we have in our heads when studying a social phenomenon and on the other hand such data as have been collected and analysed through systematic use of scientific method. This chapter introduces the building stones of this reconstruction. What characterises scientific concepts, theories, models, and data, and what role do they play in the research process?

What rules and approaches do we use when trying to understand and explain observations by help of theory, or when changing the theories as a result of new observations and ideas? The positivist tradition emphasises the alternation between on the one hand observations and the formulation of general theories, also called laws, and on the other hand testing of theories through new observations. The logic of such hypothetical-deductive reasoning, however, is mainly directed at testing theories to find out whether they are valid. It has less to say on the development of new ideas and theories. Other research logics are more focused on how to proceed in order to eliminate rivalling ideas and theories as far as possible. They build on a more pluralistic basis: that there usually are several possible explanations for and interpretations of social patterns and events.

In abductive thinking we go backwards from an observed conclusion and form premises and causal mechanisms of a kind that if they were true, the conclusion would be reasonable. Through abductive thinking we can develop different possible explanations of one and the same phenomenon. Thus, such thinking also opens up for different forms of testing of ideas through practical testing or production of data that are relevant for the testing of connections and relations.

The focus of analysis in hermeneutics is either an utterance in light of a given situation or an utterance in light of the actor who makes the utterance and wants to achieve something. In discourse analysis, we study how a given discourse order produces specific discursive practices and thus the actors’ interpretation of the world, or how subjects are constructing their identity in a given discursive context. These logics also focus on the mutual interaction between phenomenon and context. Strategic actors can change a given situation, and discursive practices can produce new frames of understanding and orders of discourse. We can say that hermeneutics and discourse analysis are concerned with the production of contextualised and situational explanations rather than with abstract generalisations that ignore time and space.

This chapter takes its starting point in the definition of research designs that I presented in the introduction: A research design is a plan for how to carry out a research project. This plan or structure has two parts. One is the purpose of the project, in other words what knowledge we want to develop about what. I distinguish between different purposes that can govern the research interest in a research project: Theory testing, theory development, theoretical interpretation, and lastly an intervention orientation. The other element in the design has to do with how we construct data to answer our questions. Here I distinguish between variable-centred and case-centred strategies as two main forms of data construction and show how they lead to different approaches in the production and analysis of data. Used for different research purposes, these two basic strategies for data construction constitute a number of design variants discussed in the next two chapters about variable-centred designs and case designs. We can also combine the main strategies for design in various ways. Depending on the purpose, such combinations are the basis of integrated designs, comparative designs, and intervention-oriented designs, all of which are introduced in the following chapters.

Variable-centred designs are often called quantitative designs. Such designs are commonly defined as building on quantitative data and on data from multiple units. Their basic property is, however, the underlying reductionism that decides how we produce and organise data in a variable-centred design. This chapter discusses how theory testing and theory development place different demands on the set-up of a variable-centred inquiry, before ending with an overview of main variants of variable-centred designs. Experimental designs, cross-sectional designs, and time series designs are discussed.

In a case study, we study social phenomena in their real context. Also, we study the case as an instance of a larger class of events or cases. It is important to understand what constitutes the case in a case study and what class of phenomena it represents. A case study needs to be focused; that is, it must be oriented towards a specific type of research purpose, and it must have a focus that is relevant for the purpose. Further, the collection of data must be structured in such a way that the collected information can be used for theoretical interpretation of the case or be relevant for the development or testing of theories. In this chapter, I discuss in detail what these requirements involve, using them to introduce variants of case designs. Discussed are designs with critical cases, congruence designs, designs for process tracing, and designs for theoretical interpretative reconstruction.

Integrated designs use different strategies for data construction in the same project. It is a complex field, and a number of different designs and methods are proposed and used. “Mixed methods”, “nested analysis”, “multimethod approach”, “comparative method”, and “qualitative comparative analysis” (QCA) are terms for variants of integrated designs that are quite different yet have in common that they use more than one strategy for data construction. I begin with a discussion of triangulation, a concept that paved the way for integrated designs by proposing the use of different data sets and analytical strategies to make analyses more robust. Triangulation focuses on arriving at the same result using different methods. Another purpose that gives reasons for the choice of an integrated design is the ambition of integrating theory development and theory testing in the same project. I discuss nested analysis as the most thorough example of a strategy that explicitly departs from this purpose. A third purpose of integrated designs deals with the development of strategies to uncover and test theories about social mechanisms. A focus on social mechanisms is typical of the literature on the multimethod approach. A main point in this literature is that the uncovering of social mechanisms requires detailed studies of one or more cases, whereas testing across cases can be made through statistical methods, by experiments, or by structured comparisons.

A comparative design involves studying variation by comparing a limited number of cases without using statistical probability analyses. Such designs are particularly useful for knowledge development when we lack the conditions for control through variable-centred, quasi-experimental designs. Comparative designs often combine different research strategies by using one strategy to analyse properties of a single case and another strategy for comparing cases. A common combination is the use of a type of case design to analyse within the cases, and a variable-centred design to compare cases. Case-oriented approaches can also be used for analysis both within and between cases. Typologies and typological theories play an important role in such a design. In this chapter I discuss the two types separately.

In the present chapter I discuss two types of intervention-oriented designs by looking at purpose and strategies in evaluation research and action research. My main argument is that comprehensive evaluations of interventions and measures must combine studies of process and effect. Measures, interventions, and reforms in society build on a more or less explicit theory of action about the effects of the measure. To evaluate the measures, we need to know something about the effects, and this can best be done through a variable-centred effect study with a control group. However, in order to develop knowledge about why the action succeeded or failed in producing the expected effects, we also need to analyse the underlying theories of action and implementation processes. Therefore, it is possible to talk about intervention-oriented designs as combinations of process-oriented and effect-oriented analysis.

As has already been stated in this book, in a design process we must find a way to link our research interest and questions with, on the one hand, general ideas and theories, on the other, a strategy for data construction. A decisive test of whether the topic is worth researching is that we are able to show its importance or relevance to others and not alone to ourselves. Thus, the first part of the design process deals with clarifying and narrowing the research purpose—together with a problem statement and possibly also detailed research questions. To do this clarification we must first delimit and justify the topic for research. Then we have to get familiar with established research knowledge about the topic in order to be able to show where our project stands in relation to this knowledge. Finally, we must introduce the theories and concepts that guide our own inquiry and clarify a frame for analysis.

The second part of the design process has to do with clarifying and concretising a strategy for data construction. With purpose and problem statement as the point of departure we have to decide on a suitable strategy for data construction. Are the purpose and the field of research such that a holistic perspective would be most suitable, meaning that we construct data as cases? Or should we use the reductionist perspective and construct data as units with variable properties? As mentioned above, the purpose may also make it relevant to choose combinations of these strategies. When we have chosen a strategy for data construction, we are ready to deal with the detailing of the research method, how to select, collect, and analyse data.