مثلث بندی بین مطالعه موردی و روش های بررسی در پژوهش حسابداری مدیریت: ارزیابی پیامدهای اعتبار

Repeated calls for validating empirical research by combining qualitative and quantitative methods have recently been made in management accounting research conducted within the positivist and functionalist paradigms. This paper provides a comparative review of management accounting research relying on triangulation between case study and survey methods with specific reference to the emerging validity implications. This research is classified into studies primarily relying on a theory testing and development logic, respectively, and works characterised by a more balanced emphasis on these logics. Our review suggests that theory-testing approaches are primarily geared towards external and construct validation and internal validation through corroboration of converging findings in line with some replication logic. By contrast, a theory development approach is typically based on internal validation efforts stemming from further probing of unexpected or inconclusive findings as a means of theory extension. An argument is advanced for combining elements of these approaches in a more balanced fashion within coherent research programmes based on multiple iterations between case study and survey methods to enable a broader range of potential validity threats to be addressed. Some general areas for further development of such a triangulation approach are also identified.

In recent years, management accounting research conducted within the positivist and functionalist paradigms has shown increasing recognition of the need to complement established quantitative methods with a greater or lesser element of qualitative, case study-based research. A number of calls for such a complementary approach, relying on method triangulation, have been made (e.g., Ferreira and Merchant, 1992, Ittner and Larcker, 2001 and Shields, 1997) and a growing number of advances in this direction can be found in empirical management accounting research. The key justification for method triangulation is typically enhancement of the validity of research findings (Brewer and Hunter, 1989, Bryman, 1992, Denzin, 1978 and Jick, 1979). However, a more systematic and in-depth discussion of validity implications associated with method triangulation has largely eluded the evolving methodological debate in the management accounting literature. A few prior attempts to contrast validity aspects associated with laboratory, survey and case study research in management accounting have been made (Abernethy et al., 1999, Birnberg et al., 1990 and Brownell, 1995). A limitation of these relatively broadly based reviews is that they primarily draw on multiple studies, each dominated by one or the other method, whilst little attention has been paid to validity issues arising from different triangulation approaches in single, or closely related studies conducted by the same researcher(s) (cf. Creswell, 1994 and Tashakkori and Teddlie, 1998).1 Further, the rationale for method triangulation as a means of avoiding potential validity threats has typically been formulated in relatively general terms based on the potential strengths and weaknesses of different methods (see e.g., Abernethy et al., 1999 and Birnberg et al., 1990). As noted by Brewer and Hunter (1989) such a priori assumptions are insufficient for making informed judgements of how potential validity threats may be reduced in specific studies. Our understanding of how triangulation between qualitative and quantitative methods may be used as a validation technique in management accounting research thus requires further development. Specifically, a systematic attempt to take stock of prior experiences with different triangulation approaches would seem warranted. The purpose of this paper is to explicate these issues, based on a comparative review of relevant management accounting research. We concentrate our review to research combining elements of (qualitative) case study and (quantitative) survey methods, which constitutes the dominant mode of triangulating qualitative and quantitative methods in the management accounting literature.2 The remainder of the paper is structured as follows. In Section 2, we describe the notion of triangulation, especially method and theory triangulation, in some detail. This leads us to propose a few criteria for classifying and assessing empirical research. In Section 3, the use of triangulation is related to the three key issues of external, internal and construct validity, commonly used to assess research within the positivist tradition, whether of a predominantly quantitative or qualitative nature (see e.g., Abernethy et al., 1999, Cook and Campbell, 1979 and Yin, 1984). In Section 4, we review a number of attempts to combine case study and survey methods in the examination of specific management accounting issues with an eye to the triangulation approach adopted and the subsequent implications for the validity issues previously raised. We conclude the paper with a discussion of the ramifications of our observations in Section 5.

Our review reveals how variations in triangulation approaches impinge on the validation of empirical findings. This contributes to nuance the rather general and unproblematic claims concerning the merits of triangulated research in the management accounting literature (see e.g., Abernethy et al., 1999 and Birnberg et al., 1990). More careful consideration should be given to the specific validity threats likely to emerge, given the choice of a particular triangulation approach and how these might be mitigated. A useful starting point for such considerations is to recognize that the types of validation efforts involved in triangulated research can vary depending on whether theory testing or development is the principal aim. Where a theory testing logic dominates (group 1 above), triangulation efforts tend towards external and construct validation through the development of hypotheses and/or measurement instrument applied in survey-based research. This is a suitable approach where prior theorizing provides reasonably well-established constructs and a solid understanding of causal relationships albeit some modifications of theory may be required depending on context-specific factors. Research may thus entail elements of a replication logic facilitating the process of analytical generalisation (cf. Eisenhardt, 1989 and Lindsay, 1995). However, internal validation is often limited to the corroboration of converging findings between different methods. Whilst some examples of qualitative methods also being used for enriching or refining causal explanations were found, researchers relying on a pronounced theory testing logic seem less inclined to rely on these for radically extending theory by elaborating alternative explanations of unexpected or inconclusive survey findings. By contrast, the studies primarily aiming at theory development (group 2 above) originate from failures to corroborate particular theoretical models or hypotheses and use triangulation as a means of extending established theory based on the examination of alternative or complementary causal relationships. Such internal validation efforts typically take the form of ex post probing of survey results. This may also enable researchers to examine potential construct validity threats emerging after the distribution of surveys (see Covaleski and Dirsmith, 1983). Whilst this gives research a more exploratory flavour, theory extensions based on further examination of inconclusive findings are likely to benefit from extensive reliance on theory triangulation. This contributes to some external validation through analytical generalisation and adds a theoretically informed explanatory element to empirical research (cf. Humphrey and Scapens, 1996 and Ryan et al., 2002). The discussion above is indicative of the critical role of the sequence in which different methods are applied in triangulated research (Brannen, 1992, Creswell, 1994 and Tashakkori and Teddlie, 1998). Although all three validity criteria may be addressed to varying extents irrespective of whether a theory testing or development logic dominates, these logics tend to be associated with a specific sequence facilitating certain types of validation efforts. For example, most studies in group 1 concentrate the use of qualitative methods to the early phases of research, whilst the studies in group 2 mainly invoke qualitatively informed explanations after the survey. As noted above the main contribution of the former approach lies in testing and corroborating causal relationships whilst the latter seems more conducive to extending theory by examining inconclusive survey findings (see also Brannen, 1992). This distinction suggests that triangulation approaches relying on closer interweaving of case study and survey methods in multiple iterations stand a better chance of addressing a broader range of validity issues (cf. Bryman, 1992 and Lee, 1991). Indeed, this is the case in research relying on a more balanced emphasis on theory testing and development (group 3 above). Whilst this group of studies includes a broad spectrum of approaches, most of them share a concern with using between-method triangulation for addressing internal validity issues pertaining to both converging and diverging (or unexpected) findings. This, in turn, gives rise to a broader range of external validation efforts combining elements of a replication and theory extension logic. It also underscores the merits of locating multiple iterations between qualitative and quantitative methods within an evolving or established research programme (Birnberg et al., 1990 and Lindsay, 1995). Specifically, this may facilitate the process of analytical generalisation back to existing and emerging theories irrespective of whether empirical findings initially converge or diverge (cf. Abernethy et al., 1999). However, this needs to be weighed against the considerable time and resources required for building up more extensive research programmes. Whilst some progress in the use of between-method triangulation has thus been made in empirical management accounting research, further development is required in a number of respects. Despite the emergence of some clearly identifiable research programmes, there would still seem to be considerable scope for elaborating triangulation approaches combining replication and theory extension logics in more equal measure (cf. Brewer and Hunter, 1989 and Lindsay, 1995). Further, as far as internal validation is concerned, qualitatively inferred explanations of inconclusive or unexpected survey findings are rarely corroborated through additional survey-based tests in the same empirical settings before theory extensions are specified (but see Covaleski and Dirsmith, 1983). Such extensions are mainly limited to analytical generalisations, with researchers primarily relying on qualitative methods in combination with theory triangulation (see also Bryman, 1992). Finally, the use of between-method triangulation for construct validation is mostly confined to instrument development and pilot testing prior to the distribution of surveys, which may testify to the complexity and costs of ex post construct validation (Brewer and Hunter, 1989). However, the latter might be potentially rewarding, especially from a theory development perspective (see Covaleski and Dirsmith, 1983). A limitation of our review is the lack of attention to methodological issues arising from the combination of between-method triangulation with theoretical perspectives grounded in diverging epistemologies (see Lee, 1991). Whilst most of the studies reviewed are based on a positivist or functionalist paradigm, embedded in more or less explicit applications of contingency theory, some attempts at triangulation with theories with stronger interpretive and social constructivist underpinnings (e.g., institutional theory) can be observed (Alam, 1997, Covaleski and Dirsmith, 1983, Hoque and Hopper, 1994, Hoque and Hopper, 1997 and Modell and Lee, 2001). This is a defensible position if one accepts that pragmatic method and theory choices may achieve a certain degree of independence from their epistemological roots (Bryman, 1992 and Hoque and Hopper, 1997). However, it has stood far from uncontested (see, e.g., Blaikie, 1991 and Sale et al., 2002). Applying validity criteria originating from the positivist research tradition would nevertheless seem acceptable as the studies cited above are clearly concerned with suggesting plausible causal relationships. Further, the kind of theorizing invoked persistently testifies to concerns with analytical generalisation. Although we recognize the critique that may be levelled at our position debating subsequent epistemological ramifications is beyond the scope of this paper.