an d p d tio 12 t lik are white. In social sciences most hypotheses are probabil- study research and to test theories with case studies. At the moment, the debate about these applications of case study statistically analyze quantitative data from a large number surveys we can increase our confidence in the direction of and by including rival explanations in the analysis (control play an important role in theory testing. First of all, let ARTICLE IN PRESS us consider theory testing for deterministic theories. In the approach of Dul and Hak (2008), the case study test simply means comparing the observed values of the 1478-4092/$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.pursup.2007.09.012 E-mail address:
[email protected] istic in nature however, which means that the hypothesized association between the independent variable and the dependent variable is expected to be true in general. Under this assumption the analysis of variables in a single case has no implications for the validity of the hypothesis. This does not mean however that case study research is unsuitable for theory testing. In this comment I complement the contributions of Dubois and Araujo and of Dul and Hak by briefly presenting some alternative ways to apply case variables). In addition, we might include variables that are expected to mediate the relationship between the indepen- dent and the dependent variable. These mediating variables reflect the causal mechanisms that explain how and why X causes Y (Van de Ven, 2007). Although survey research can include mediating vari- ables, case study research is capable of studying causal mechanisms in much greater detail (Yin, 2003). This capability is mostly used for the purpose of theory building, but the study of causal mechanisms can also ‘crucial case’ (Gerring, 2007) is similar to the observation of a black swan, which would reject the claim that all swans causality by having multiple observations over time (to capture the temporal sequence between cause and effect) sufficiency of a condition for a certain outcome. Such a Journal of Purchasing & Supply M Notes an Case study research: Some other a Ferdinan Department of Management of Technology and Innova Received 4 June 2007; received in revised form Keywords: Case study research; Survey research; Casual mechanisms; Mos In the field of purchasing and supply management the main application of case study research is theory building, whereas survey research is the dominant approach for theory testing. This is also true for management research in general. Case study research can also serve other purposes however besides theory building. Dubois and Araujo (2007) mention that case studies can also be used for the purpose of theory testing, but their article mainly focuses on theory building. In their comment, Dul and Hak (2007) complement Dubois and Araujo by discussing in greater detail the possibility to test theory with case studies. Dul and Hak argue that case studies are very suitable for testing deterministic hypotheses, whereas surveys are preferable for testing probabilistic hypotheses. Deterministic hypotheses formulate relationships be- tween variables that must hold in all cases for the theory to be true. This means that an individual case can decisively disconfirm the claim about for instance the necessity or the of cases. If the size and the sign of the association (e.g. a correlation or regression coefficient) confirm the hypoth- esis, this does not mean however that evidence is found about the direction of causality between X and Y, i.e. it remains unclear whether X causes Y or vice versa. With agement 13 (2007) 210–212 Debates plications besides theory building Jaspers n, RSM Erasmus University, Rotterdam, Netherlands September 2007; accepted 12 September 2007 ely and lease likely cases; Process-tracing research mainly takes place in political science (see for instance George and Bennett, 2005; Gerring, 2007). This is unfortunate given the benefits that these applications could bring to the field of purchasing and supply management. 1. Process-tracing cases A simple probabilistic theory might specify why it is expected that an increase in X is likely to result in an increase of Y. To test the hypothesis that a positive association exists between X and Y, survey research would www.elsevier.com/locate/pursup ARTICLE IN PRESS Sup independent variable and the dependent variable with their expected values. This test can be based on a visual inspection of the data and does not require statistical techniques and neither does it involve a within-case analysis to observe causal mechanisms. If the deterministic theory also specifies through which causal mechanisms the independent variable causes the dependent variable, this provides the possibility for much more convincing tests in what George and Bennett (2005) label process- tracing cases. Process tracing involves the study of the causal mechan- isms within individual cases and compares the observed mechanisms with the precise predictions about how and why a certain independent variable influences the depen- dent variable. This means that process tracing really analyzes the direction between cause and effect. In addition, process tracing involves a very strong test because the theory only gets confirmed if the observed mechanisms and processes perfectly match the expected mechanisms (George and Bennett, 2005). Hence, these process-tracing tests of individual cases provide greater insights about the validity of the theory than case studies that would only compare the values of X and Y. 2. Most likely cases and least likely cases Secondly, causal mechanisms can also play a role in the testing of probabilistic theories with case studies. Although it is by definition impossible to decisively disconfirm a probabilistic theory with a crucial case (Dul and Hak, 2007), case studies can still have strong implications for our confidence in such theories. Gerring (2007) for instance advises to select cases on theoretical grounds in order to maximize their ‘degree of crucialness’, i.e. the extent that they can confirm or disconfirm the hypothesis. Consider again our earlier example of a theory that specifies how and why X is expected to have a probabilistic effect on Y. For this theory a so-called most likely case would be a case with an extreme value of X, i.e. high or low. This case is therefore most likely to show how X influences Y. The observation in such a case that X does not influence Y or that X influences Y in the opposite direction would challenge our confidence in the theory (George and Bennett, 2005). Conversely, a least likely case would be a case for which the observed value of X is high (low) and for which rival explanations would predict low (high) values of Y, i.e. values opposite to the expectation based on the value of X alone. If a within-case analysis reveals that X contributes to a high (low) value of Y, then this would increase our confidence in the theory (George and Bennett, 2005). In sum, the analysis of causal mechanisms within theoretically selected cases can have strong implications for our confidence in the validity of probabilistic theories. Of course our confidence in the validity of the theory grows F. Jaspers / Journal of Purchasing & stronger or weaker if we are able to replicate case study findings. 3. Pathway cases Above we have seen the importance of the causal mechanisms that explain how and why a cause is expected to result in an effect. Theories in social science are often not very precise in their causal explanation however, i.e. many theories are rather ambiguous regarding the inner workings of the hypothesized causal effect (Gerring, 2007). A causal factor might for instance result in the same effect through many different causal mechanisms (equifinality). This makes the precise definition of theories—as required for process-tracing tests—very difficult. In relation to this imprecision about causal mechanisms between cause and effect, Gerring (2007) introduces the concept of the pathway case. The aim of this type of case study is to ‘elucidate’ the causal mechanisms for theories about which we already know that a significant association exists between the independent variable and the dependent variable. Hence, insights from pathway cases do not result in the reformula- tion of hypotheses, but rather in the reformulation of the theory itself (Gerring, 2007). In other words, pathway cases can be said to enhance the quality of explanation as a result of their focus on ‘‘the cogs and wheels behind the regression coefficients’’ (Davis and Marquis, 2005, p. 341). Pathway cases are those cases for which we have reasons to believe that the causal effect is as strong as possible. An example would be a case characterized by an extreme value of the independent variable, by a value of the dependent variable that is to be expected for this extreme value, and by the absence of any rival explanations. 4. Conclusion For the purpose of theory testing with case studies or surveys, Dul and Hak describe two ends of a continuum. On one extreme, deterministic theories can best be tested with single cases (crucial cases). On the other extreme, surveys are best positioned to test probabilistic theories. Both methods only analyze the values of the independent and the dependent variables and they do not analyze the causal mechanisms that operate within individual cases. In this comment I described other ways to test theory with case study research: process-tracing cases (for deterministic theories) and most likely and least likely cases (for probabilistic conditions). Finally, pathway cases are able to improve our understanding of the causal mechanisms for well-established associations between variables. These alternative applications of case study research fully exploit the potential to analyze causal mechanisms within cases. This might significantly increase our understanding of theories and might also provide important insights for managers. Finally, I would like to remark that our unfamiliarity with alternative applications of case study research might have resulted in too many case studies with a theory- ply Management 13 (2007) 210–212 211 building objective. I believe that it would have been perfectly possible for many case studies in (purchasing and supply) management research to formulate hypotheses up front and that the only reason to label these studies ‘exploratory’ or ‘inductive’ is that the number of cases is too small to apply statistical techniques. This is an incorrect line of reasoning however, because case study research and survey research each have their own strengths and weaknesses and should be viewed as highly comple- mentary (George and Bennett, 2005). Hence, once we for instance recognize the role that case studies can play in the (dis)confirmation of probabilistic hypotheses, case studies that otherwise would have been presented as ‘theory building’ might be subjected to strict theoretical selection criteria and might contribute significantly to our under- standing of purchasing and supply management. References Davis, G.F., Marquis, C., 2005. Prospects for organization theory in the early twenty-first century: institutional fields and mechanisms. Organization Science 16 (4), 332–343. Dubois, A., Araujo, L., 2007. Case research in purchasing and supply management: Opportunities and challenges. Journal of Purchasing & Supply Management 13 (3), 170–181. Dul, J., Hak, I., 2007. To quantify or to qualify: That’s not the question. Journal of Purchasing & Supply Management 13 (3), 207–209. Dul, J., Hak, T., 2008. Case Study Research Methodology in Business Research, Butterworth-Heinemann, Oxford. George, A.L., Bennett, A., 2005. Case Studies and Theory Development in the Social Sciences. MIT Press, Cambridge, MA. Gerring, J., 2007. Is there a (viable) crucial-case method? Comparative Political Studies 40 (3), 231–253. Van de Ven, A.H., 2007. Engaged Scholarship; A Guide for Organizational and Social Research. Oxford University Press, New York. Yin, R.K., 2003. Case Study Research; Design and Methods, third ed. Sage, Thousand Oaks, CA. ARTICLE IN PRESS F. Jaspers / Journal of Purchasing & Supply Management 13 (2007) 210–212212 Case study research: Some other applications besides theory building Process-tracing cases Most likely cases and least likely cases Pathway cases Conclusion References