استفاده از مدل سازی حداقل مربعات مشتقات جزئی در تحقیقات حسابداری

Partial least squares (PLS) is an approach to structural equation modeling (SEM) that is extensively used in the social sciences to analyze quantitative data. However, PLS has not been as readily adopted in the accounting discipline. A review of the accounting literature found 20 studies in a subset of accounting journals that used PLS as the data analysis tool. PLS allows researchers to analyze the measurement model simultaneously with the structural model and allows researchers to adopt more complex research models with both moderating and mediating relationships. This paper assists accounting researchers that may be interested in adopting PLS as an analysis tool. We explain the benefits of using PLS and compare and contrast this analysis approach with both ordinary least squares regression and covariance-based SEM. We also explain how the PLS algorithm works to derive estimates for the measurement and structural models. To further assist researchers interested in using PLS, we offer guidelines in the development of research models, analysis of the data, and the interpretation of these results with PLS. We apply these guidelines to the accounting studies that have used PLS and offer further recommendations about how researchers could apply PLS in future accounting research.

Partial least squares (PLS) path modeling is an approach to structural equation modeling (SEM) that has been in use for many years in the field of psychology and the social sciences, including many business disciplines such as marketing (e.g. Fornell and Bookstein, 1982) and information systems (e.g. Chin, 1998b). Studies utilizing PLS as a method for model estimation and testing routinely appear in leading information systems journals, as well as in leading multi-disciplinary business journals. However, despite frequent use of this technique in other business disciplines, the accounting discipline has been slower in its general acceptance of PLS and other SEM modeling techniques. The reluctance to use PLS in accounting research perhaps may be due to a lack of understanding of PLS's benefits and applicability in accounting research. PLS, like other SEM techniques, enables a set of relationships among one or more independent variables and one or more dependent variables to be examined in a comprehensive model. Whereas traditional regression may require separate regression equations to analyze each hypothesized relationship, PLS allows the system of equations to be analyzed simultaneously. In addition, PLS and other SEM techniques allow for the analysis of both directly measured variables and latent variables. Because PLS is closely associated with the analysis of latent constructs, it has been frequently used in survey-based research. However, PLS is not methodologically tied to surveys and has been used with data collected via other mechanisms, such as experiments (Feldman et al., 1998) and archival data (Ittner et al., 1997). Our goal in this methodological note is to help eliminate barriers that might prevent accounting researchers from using PLS. Accounting researchers such as Hall et al., 2005 and Blanthorne et al., 2006 have recognized the potential benefits of using more sophisticated data analysis techniques such as SEM on traditional accounting data sets. To achieve this goal, we first provide a non-technical description of how PLS derives statistical parameters. Next, we provide guidelines for researchers to consider when conducting research to prevent common mistakes with PLS which could negatively impact the results, analysis, or possibility of publication. Finally, we examine selected accounting journals to identify the frequency of use of PLS, as well as the extent that the existing research conforms to the guidelines presented in this study. Not surprisingly, our review finds relatively infrequent use of PLS in published accounting research; nevertheless, the review does demonstrate that accounting research using PLS is generally consistent with our guidelines. Finally, we conclude with a discussion on limitations of this methodological note and suggest areas of future research.

7.1. Contributions In this paper, we presented an overview of how PLS works, general guidelines for PLS usage, and a discussion of the use of PLS in recent accounting literature. The primary contribution of this paper is methodological with the goal of clearly identifying guidelines for the appropriate use of PLS. By explaining the underlying PLS algorithm, researchers more familiar with traditional regression can better understand what PLS is and why and when PLS may be appropriate. By documenting the appropriate statistics to report when using PLS, accounting researchers will have a clearer understanding of how to use PLS in their research. A secondary contribution of this paper is the documentation of prior usage of PLS in the accounting literature and evaluating the conformance of these papers to best practices in PLS. As shown in Table 9, although the usage of PLS in accounting has been limited, much of this usage is in conformance to best practice guidelines. These studies identified in Table 9 can serve as possible examples for researchers interested in applying PLS. 7.2. Future research and limitations While many of the example usage of PLS in the accounting literature has been survey-based, we would like to reiterate that the use of PLS is not limited to surveys. PLS is primarily used in the evaluation of latent constructs, and the data can be based on archival sources, experimental results, surveys, etc. The additional model testing capabilities of SEM and PLS can be used to further extend existing research and provide support for accounting research and in particular archival accounting research (e.g., Ittner et al., 1997). Perhaps future studies can elaborate on these additional benefits given the wide use of first generation regression techniques employed in accounting research. Another area of future research is the problem of construct misspecification for formative and reflective constructs. While this study focused on construct misspecification in PLS studies, construct misspecification is a potential problem in all research involving latent constructs. Therefore, future research could investigate the extent of construct misspecification in accounting literature and possible remedial methods to address the problem. A final area of future research would be the further analysis of covariance-based SEM studies in the accounting literature. Future research could investigate the extent of usage of covariance-based SEM in accounting, as well as advance SEM applications such as latent growth modeling (Meredith and Tisak, 1990) in the accounting context. This paper has limitations that can be addressed in future studies. First, this paper examined a subset of accounting journals and found limited use of PLS in the sample frame. A more comprehensive journal search may reveal a wider usage rate of PLS or reveal additional problems with PLS not identified in the current journal sample. Another limitation of this study is its emphasis on only component-based SEM modeling. In this study, we did not examine the extent of covariance-based SEM techniques like LISREL in the accounting literature or focus on problem areas associated with covariance-based techniques. We purposefully chose to focus on the use of PLS since it has more in common with regression, which is commonly used and accepted within the accounting literature. A final limitation of this study is the lack of definitive guidelines on the issue of power and effect size as related to PLS. Although best practice guidelines recommend that power and effect sizes be calculated as part of a PLS analysis (e.g. Marcoulides and Saunders, 2006), scant practical guidelines are available in the PLS context to guide researchers in this area. PLS is a data analysis technique that has been widely used in other business disciplines. We hope that this methodological note will assist accounting researchers in understanding what PLS is, when PLS is appropriate, and how to properly incorporate PLS usage in their research.