Mixed-methods design in practice

Since the building of new knowledge concerns a multidimensional and complex socio-economic process that combines events, mechanisms, and actors, there has been a need to find a mix of methods that is carefully tailored to this case study (cf. Tykkyläinen 2015, 42). The logic continues the idea of Yeung (2003), whereby different kinds of data and methods of analysis have been deployed which connect the thesis under a mixed-methods design (Creswell 2003). The simple advantage of mixed methods design is that it enables the use of multiple methods when one method is insufficient for giving an answer to the research question (Morse & Niehaus 2009, 9) which is here understood as a practical reason to deploy mix-methods. In the thesis, the whole research orientation is based on mixed-research design starting from processual proceedings in regard to methodologies. The mixed-methods more specifically are understood to refer to the practices of doing research in the articles, which have required the mixing of quantitative and qualitative data and analytic techniques. As has been the case throughout the methodologies, the designing of mixed-methods has been processual (Table 2).

Table 2. Method and data by article (I–V)

Mixed-method design Original data sources: amount and data category I Explanatory triangulation:

Statistical descriptions explained by contextual literature and additional documents

Statistics: 8 economic, 6 social

Documents: 7 histories, 6 reports, 9 media, 2 strategies, 4 data printouts

Documents: 6 histories, 5 media, 2 strategies, 1 legislation

Documents: 2 histories, 8 reports, 8 media, 4 strategies

IV Basic triangulation:

Documents: 1 history, 7 reports, 1 media Interviews: 8 local policymakers, 11 in-migrants (also 9 entrepreneurs)

V Embedded triangulation:

Interviews analysed as narratives that are supported by statistical data, contextual literature, and additional documents

Interviews: 8 local policymakers Statistics: 7 economic

Documents: 2 reports, 1 media, 1 legislation

In addition to practicality, the mixing of methods deployed in a processual manner is seen as relevant because it is a way to strengthen the construct validity of the case study by using multiple sources of evidence (see Yin 2003, 34–35, 85–97) through the process. In this thesis, the evidential sources are formulated through three main data

sources: statistics, documents, and interviews, which include different types of data categories (Table 2). During the process, the construct validity is seen as improved by joining different data sources together, which in practice meant not only comparing one data with another but most often explaining an initial data by relating and matching it with multiple data gathered from different sources. The initial data are used for the selection of the specific types of changes that are explained by the other data from the perspective of a research question; together they are seen as a way to assist in focusing on data which are relevant to answering the research question.

As the aim has been to find causal explanations, attention concerning internal validity is required, which refers to the correctness of the causal explanations. In the analysis here is adopted a procedure by Yin (2003, 34–36) in a format where a form of time-series analysis and explanation building are mixed. As a starting point, a timeline had to be described and used to observe the chronologically proceeding phases which reveal the interesting turning points where to focus. As is typical for complex time series, the explanations of the outcomes should also include the comparison of different types of explanations and their outcomes (Yin 2003, 124–125). Especially when some data has led to one causal explanation and other data to a different one, this has served as a caution to rethink the reasoning behind different conclusions. The explanations and their reasoning have been checked, complemented with new data, and the analysis was continued until a satisfactory synthesis of explanations was conceivable.

Explanation building (Yin 2003, 120–122) is deployed here to strengthen the causal explanations based on the empirical data with former propositions concerning the causalities that have been used to explain the socio-economic changes in Finland and on the different resource peripheries as well as evolution and resilience of different localities in general.

Articles I–III are executed as a mix of explanatory triangulation (Table 2), where the quantitative data have been taken as initial data; these are explained through qualitative data, and triangulation is utilised to validate the explanations by comparing or contrasting the different types of data (see Creswell & Plano Clark 2007, 62, 71–75).

Comparing is here understood as a way to validate the similar type of causalities, which requires understanding the reality as objectively existing. Contrasting, instead, is understood to refer to validity in a sense that different types of understandings of reality deepen the knowledge regarding causalities by their synthesis, but they are not themselves comparable. As the purpose of the quantitative statistics data is to describe the booms and busts from different socio-economic perspectives, they give a skeleton for what needs to be explained. For some parts, the causalities of a certain boom or bust could be pre-explained and validated by other statistics, but the main explanations have required qualitative documents. Article IV represents the most typical formation of triangulation, where the interpretation is based on the synthesis of quantitative and qualitative analysis (Creswell & Plano Clark 2007, 63). To some extent the article also includes explanatory parts, but the emphasis is on finding causalities and interpretations by linking statistics and documents; this analysis is complemented by selective notes from the interview data. Article V follows a kind of variation of embedded triangulation. It is embedded because one type of data is clearly an initial one, and the other type has been used only for supportive purposes (Creswell & Plano Clark 2007, 67). In article V, the interviews as initial data were analysed as narratives that are visually supported by the backgrounding statistical data. The triangulation concerning the validity is executed in a way that the narratives are contrasted with the previous explanations in articles I–IV. The mixing of different kinds of data and types

of analysis was done for the purposes of practicality and validity, but also because it has been seen as a way to diminish the limitations of knowing reality. The different kinds of knowledge sources and thence the improvements in knowledge were gained by using different types of complementary data.

As with any research, this case study also should be implemented in an open way so that any other researcher or other critical evaluator can observe where, when, and how the explanations have been formulated. Especially in case studies where the documentation of data has been done poorly, the reliability has been questioned because they have been seen as prone to errors and biases caused by disordered data (see Yin 2003, 37–39). Here the reliability of the case study has been strengthened through detailed descriptions of the original data sources and presentation of the data next to explanatory information. This is seen as especially important because the causalities and explanations have been gathered from many different types of data sources, which could have led to disordered data gathering and thence biased explanations. All the original data sources are mentioned in the articles by using references according to the type of data. The statistics and documents are typified as secondary data reformulated for analytic purposes. These secondary data are referred with the same protocol as any other references: author information (if mentioned), title of data source and publisher (if mentioned), and date of access for electronic data.

Most of the statistical data sources are reshaped for graphical analytic purposes, which are visually presented in the vicinity of the text where the causal explanations are presented. The interviews are seen as primary data with two purposes: in article IV the advancing role of the other data sources and in article V the initial role through which the constructions of reality are produced. In article IV, the reference to the interviewees as a group has been made in the vicinity of the information that mainly describes the metadata about the interviewees and lists their preferences as entrepreneurial migrants. In article V, the constructions of reality are produced by the collection of the narratives of individual interviewees, whereby a researcher was more responsible for the interpretations. For this reason there was a need to present quotes from the original interview data and thus make the interpretation process more visible to critical readers.

4 DISCUSSION AND TESTING OF