Chapter 5. Research Methodology
5.2 Methods of Case Study Selection using ACI
An in-depth, narrative approach to the research limits the number of case studies to be investigated on practical grounds. The following section sets out the research’s methodological approach to choosing case studies. It discusses the issues and rules for qualitative case study selection and then turns to apply these to the ACI framework. The case study selection process is then described.
Methodological considerations
Carefully designed comparative research enables intellectual rigor in qualitative research.
However, as King, Keohane and Verba illustrate, the structure by which case studies are selected is central to the success of a qualitative study. They argue that “poor case selection can vitiate even the most ingenuous attempts, at a later stage, to make valid causal inferences”
(1994, p. 115). King, Keohane and Verba, followed by Denters and Mossberger (2006), outline the methods in designing comparative study selection. They both agree that random selection of case studies is difficult, particularly in the social sciences. In intentional or purposive sampling of case studies they argue “there are two basic options available: selection on the dependent variable and selection on the independent variable”. The dependant variable(s) are also known as 'outcome variable' (e.g. DHC delivered), and the independent variable(s) are also known as the explanatory variables or observations, with the ‘key’ independent variable the ‘cause’ of the dependant variable. By holding some independent variables constant and varying the remaining independent variables, you come to understand what influences the dependant variable.
Both papers are adamant that selecting case studies on dependant variable can lead to inability to test causal relationships between factors and variables in the case studies to be compared. If all case studies are successful in delivering DHC, and you conclude that x-variable caused the success, you cannot test for the absence of x-variable leading to failure.
65 King, Keohane and Verba (1994) provide further points regarding variable selection which are helpful. Firstly that “selecting observations for inclusion in a study according to the categories of the key explanatory variable causes no inference problems" (p. 137). Secondly that “the best 'intentional' research design selects observations to ensure variation in the explanatory variables without regard to the values of the dependant variables” (p. 140). And finally that, though the research may look at a wide range of influences or observations in each case study, it is appropriate to control selected variables, to seek out the variables with the most substantial causal relationships to the dependant variable.
Kantor and Savitch also sketch out limitations that restrict case study selection; firstly that they must have some similarities to enable like for like comparison, secondly that language can be an important factor in understanding, thirdly that comparable data should be available, and finally that conceptual parameters should be similar, e.g. the concept of ‘regional’ (2006). The challenge of conceptual language aspect is also covered in other papers but can be overcome with careful attention.
To summarise, the research design should identify the dependant variable, the outcome of the causal relationships. Selection of case studies for independent variables or even key independent variables causes no inference problems. Within those case studies selected, a wide range of potential influences should be examined. It is appropriate to hold certain influences constant in order to better understand causality between influences and success.
Dependent and independent variables in ACI
This section applies the above case study selection guidelines to the preferred conceptual framework, Actor-Centered Institutionalism (ACI) described in Chapter 4. ACI proposes that, in reference to a defined policy issue in a specific institutional setting, policy outcomes are influenced by actors interacting according to actor constellations in a defined mode of interaction (Scharpf, 1997).
For this research, the policy outcome is the delivery of DHC. Therefore, building on research design theory, the research is structured with delivery of DHC as the dependant variable – the outcome for which causal factors and relationships are sought. Case studies where DHC was sought by actors, but not delivered, are included in the case study selection; effectively this ensures that the research is not selecting case studies based on the dependant variable. This enables testing of causal relationships between explanatory variables, a core component of comparative research theory as described above.
As the selection of case studies based on independent variables causes no inference problems, and ACI provides a structure which can cope with variety and complexity of causal influences.
ACI concepts are employed to determine which independent variables to hold constant through the case study selection process. In ACI, there are five components which drive policy outcomes; the following text addresses how they will be considered in case study selection.
66 First, a particular institutional setting. To focus on network governance, the research selects for case studies with both public and private actors within them. It excludes case studies which are only private or only public sector driven. Secondly, a specific policy problem: DHC should be an environmental policy goal in each case. The research selects for case studies where the environmental benefits are a publicly stated goal and excludes case studies where environmental concerns are not mentioned. Third, specified actors. The research selects for case studies where planning institutions are involved and excludes case studies where they are not involved in the delivery of DHC. Fourth, actor constellations. This component is collective picture of actor perceptions about the policy problem. Therefore it is not a variable and the research cannot control case studies based on this variable.
Fifth, the mode of interaction. In network governance, actors are relatively stable, non-hierarchical, interdependent but autonomous, and interact through negotiations. The research excludes case studies with explicitly horizontal connections between major actors to guarantee a level of network governance patters in each case; this is similar to selecting for case studies with both public and private actors. As described in Chapter 4, the research adds a further component within the mode of interaction on the success of governance networks in the form of relational characteristics. Relational characteristics are proposed as the patterns and quality of interactions among purposeful actors. The research does not propose to select case studies on this component. A summary of these implications for the case study selection process is provided in Table 10.
ACI Component Research Approach Case Study Selection DHC as an environmental
policy goal
Static independent variable All cases have DHC as an environmental policy goal Institutional Setting Independent variable Cases have both public and
private actors
Actors Independent variable Planning activity in support of
DHC as a static independent variable
Actor Constellations Determined by problem and actor set
Table 10: ACI and the selection of case studies
Case study selection
A literature review was unable to locate an existing review of case studies of urban DHC in a comprehensive or rigorous way across the globe. Therefore a wide range of sources of information have been used to establish a ‘long list’ of case studies: researcher’s industry knowledge, global energy company websites, World Energy Council and International Energy Agency publications, Google searches, searches of academic journals, and energy statistics published by individual countries. These generated the long list of case studies in Table 11.
67
City Country Dependant
Variable Amsterdam The Netherlands Delivered
Barcelona Spain Delivered
Berlin Germany Delivered
Burlington USA Not Delivered
Cambridge USA Delivered
Chicago USA Not Delivered
Copenhagen Denmark Delivered
Dubai UAE Delivered
Frankfurt Am Main Germany Delivered
Hamburg Germany Not Delivered
Harbin China Delivered
Helsinki Finland Delivered
London (Southwark) UK Not Delivered
London (Kings Cross) UK Delivered
Lerwick UK Delivered
Miami USA Delivered
Paris France Delivered
Portland USA Not Delivered
Reykjavik Iceland Delivered
San Diego USA Delivered
Seoul South Korea Delivered
Singapore Singapore Delivered
Southampton UK Delivered
Toronto Canada Delivered
Victoria, BC Canada Not Delivered
Vienna Austria Delivered
Table 11: Long list of cities with DHC systems
In practice, selecting case studies according to the dependent variable and holding static two independent variables tends to exclude the following groups of case studies. Case studies where long-standing tax regimes or the physical climate encouraged DHC on economic grounds were not included as for those situations DHC was not a messy policy problem and network governance patterns were not observed. Similarly, case studies in countries where DHC had been practiced for many years and was no longer perceived as a environmental policy goal but simply standard practice were not included. Case studies in non-urban or industrial settings, and DHC systems installed before approximately 1980 which have not been significantly expanded or upgraded in the intervening years were also not included; the significant structural changes to both urban policy (governance patterns) and energy markets (liberalization) since that time tended to signify that non-hierarchal governance patterns were absent.
Further to the above ACI components which served as independent variables, the following influenced case study selection. First, the availability of information in English. The research avoided case studies were little or no desktop information was available in English; this was a
68 practical decision which led to the absence of Asian examples from the research scope.
Second, access to actors. Due to the need to understand organisation and individual perceptions, it was important that the research be able to interview actors involved in the cases.
Therefore the research focused on fairly recent case studies. Access to actors did remain a data concern throughout the research, as discussed below.
Applying the case study selection criteria established above to the long list of case studies in Table 11, the following case studies were chosen for detailed study.
1. Elephant and Castle, London Not delivered 2. Districlima 22@, Barcelona Delivered 3. Deep Lake Water Cooling, Toronto Delivered
4. Burlington, Virginia Not delivered
5. Lerwick, Shetland Delivered
To summarise, the research design employs ACI to inform the case study selection, being mindful of theoretical and methodological considerations of comparative case study research.
Both case studies where DHC has been delivered and where it has not been delivered are chosen. Selection of case studies for the actor component causes no inference problems. It is appropriate to hold certain variables constant in order to better understand causality between influences and success; these variables were selected by employing concepts in ACI. Case study selection holds static a key component – ACI's policy goal (DHC as an environmental policy goal) - and excludes cases studies where DHC is delivered on financial grounds alone.
Alongside the selection for cases with planning organisations actively involved, this focuses the case studies to situations which address the research questions.
In support of the comparative methodology it is helpful to standardise terminology about government actors. For the purposes of this thesis, local government or municipal government indicates the lowest form of state organisation to pro-actively manage an urban area (e.g.
London Borough of Southwark). Metropolitan government indicates the next step up within an urban context; this can be a collective of many organisations (e.g. Barcelona Metropolitan organisations, Greater London Authority). State government indicates a large sub-national level of government which can have varying degrees of autonomy and legislative power (e.g. State of Vermont, Generalitat de Catalyna, Scottish Government, Province of Ontario). National government indicates the highest level of government of a nation-state, with implications for citizenship and controlled borders.