Was leistet empirische Politikforschung?

Norman Blaikie ist Soziologieprofessor in Malaysia und Australien. Er gilt weltweit als der Spezialist, der sich mit der sozialwissenschaftlichen Forschung beschäftigt hat, um für Studierende und Forschende die Frage zu beantworten, was Sozialforschung leistet, warum man sie macht, und wie sie zu geschehen hat.
Ich bringe hier sein mustergültig kurzes Manifest, das er zur Jahrtausendwende publiziert hat, im Original. Man kann es eigentlich “eins-zu-eins” auf die empirische Politikforschung übersetzen.

Claude Longchamp


A Manifesto for Social Research

1 Social research is about answering research questions.

2 Three types of research questions can be asked: ‘what’, ‘why’ and ‘how’.

3 All research questions can be reduced to these three types.

4 Social research will also address one or more of the following objectives: exploration, description, understanding, explanation, prediction, Intervention (change), evaluation and impact assessment.

5 ‘Why’ questions are concerned with understanding or explanation. ‘How’ questions are concerned with intervention. All other objectives involve the use of ‘what’ questions.

6 Hypotheses are possible answers to ‘why’ and some ‘how’ questions. They are normally expressed as Statements of relationships between two coneepts. Hypotheses direct the researcher to collect particular data.

7 ‘What’ questions do not require hypotheses. Nothing is gained from hazard-ing an answer to a question that simply requires research to produce a description.

8 Research questions are answered by the use of four research strategies: the inductive, deductive, retroductive and abductive.

9 The major characteristics of the research strategies are as follows: the in-ductive strategy produces generalizations from data; the deductive strategy tests theories by testing hypotheses derived from them; the retroductive strategy proposes causal mechanisms or structures and tries to establish their existence; and the abductive strategy generates social scientific accounts from everyday accounts.

10 When a research project includes a variety of research questions, more than one research strategy may be required to answer them.

11 Because research strategies entail different ontological and epistemological assumptions, they may only be combined in sequence.

12 Hypotheses are used mainly in the deductive research strategy as part of the process of testing theory. While the testing of hypotheses commonly involves the use of quantitative methods, it need not do so. The deductive strategy can also use qualitative methods, in which case hypothesis testing is more in terms of a discursive argument from evidence.

13 The abductive research strategy may use hypotheses in the course of gener-ating theory, but in a different way to the deductive strategy. These hypotheses are possible answers to questions that emerge as the research proceeds. They are used to direct subsequent stages of the research.

14 The hypothetical models of possible causal structures or mechanisms that are developed in the retroductive research strategy are not hypotheses. The researcher’s task is to establish whether a postulated structure or mechanism exists and operates in the manner suggested.

15 Social science data normally Start out in the qualitative form, in words rather than numbers. They may continue in this form throughout a research project or be transformed into numbers, at the outset, or during the course of the analysis. Ultimately, research reports have to be presented in words. When numbers are used, they need to be interpreted in words.

16 The use of tests of significance is only appropriate when data have been generated from a probability sample. These tests establish whether the characteristics or relationships in the sample could be expected in the population. Tests of significance are inappropriate when non-probability samples are used, and are irrelevant when data come from a population.

17 As methods of data collection and analysis can be used in the Service of different ontological assumptions, there is no necessary connection between research strategies and methods.

18 Methods of data collection can be combined, in parallel or in sequence. However, it is only legitimate to combine methods in parallel when they are used with the same or similar ontological assumptions. That is, data generated in the Service of different ontological assumptions cannot be combined, only compared. It is legitimate to combine methods in sequence, regardless of their ontological assumptions. In this case, it is necessary to be aware of the implications of switching between assumptions.

19 Case studies are neither research designs nor methods of data collection. They constitute a method of data selection and, as such, require particular procedures for generalizing from the results produced.

20 The results of all social research are limited in time and space. Hence, making generalizations beyond a particular time and place is a matter of judgement. While quantitative data from a probability sample can be statistically gen-eralized to the population from which the sample was drawn, this type of research is in the same position as any other when it comes to moving beyond that population.

Norman Blaikie: Desining Social Reserach, 2000