STRUCTURAL EQUATION MODELING IN MANAGEMENT ACCOUNTING RESEARCH: CRITICAL ANALYSIS AND OPPORTUNITIES

10 INTRODUCTION

In new years, several papers in leading accounting journals have speculated upon the future directions of management accounting research, including the forms of theoretical modeling and data analysis that can be used [see for example, Shields, 1997; Atkinson et al., 1997; Fisher, 1995] Other papers highlight flaws in the application of regression-based statistical courses commonly used in management accounting research and identify shortcomings in theoretical modeling [Hartmann and Moer 1999; Shields and Shields, 1998] Within this general environment of criticism and uncertainty, a certain number of writers [Shields, 1997; Shields and Shields, 1998] have called upon management accounting researchers to make greater use of structural equation modeling (SEM) to provide simultaneous experiments of measurement reliability and structural relations, which may rout some of the limitations that have been horizontaled at the way that management accounting has used more traditional statistical techniques.

There are three objectives underlying this paper. The first objective is to demonstrate more [i]or[/i] less advantages that latent variable SEM may tender over the well-established techniques of multiple regression analysis and path analysis to advance the theoretical unravelling of management accounting. The next to the first objective is to undertake a critical review of empirical management accounting papers that utilize SEM to determine shortcomings, limitations, and opportunities. new papers [for example, Hartmann and Moer 1999] that highlight flaws in the application of regression-based analyses in management accounting, point to the importance of considering methodological shortcomings of SEM The final objective is to identify ways in which management accounting researchers can use SEM to advance knowledge in the discipline.

This paper is structur as tread in the steps ofs In the following section, the advantages that SEM may clinch as a statistical method are demonstrated. a certain quantity of limitations of multiple regression and path analysis that may hindrance their suitability for many impressed signs of management accounting research are considered. This is followed by the agency of a review of published management accounting research that utilizes SEM This take a view of examines the areas in which a certain quantity of management accounting research has failed to consider lock opener principles relating to the use of SEM It also examines the stamp of research problems where SEM has been used. In the final section, conclusions and suggestions for improving the application of SEM in management accounting research are not absented and the potential for SEM to advance the literature in management accounting is assessed.

20 MOVING FROM REGRESSION ANALYSIS TO PATH ANALYSIS TO SEM

SEM is a statistical technique that allows the simultaneous analysis of a series of structural equations. It is particularly useful when a pendent variable in one equation becomes an independent variable in another equation [Hair et al., 1998] However, there has been a certain quantity of confusion and inconsistency as to what constitutes SEM Hoyle [1995] defines SEM as "a comprehensive statistical approach to testing hypotheses about relations among observ and latent variables."1 Bollen and drawn out [1993] note that the use of the confine structural equation modeling is repeatedly restricted to classical econometric modeling, on the contrary extended their definition of SEM to include general path analysis. It have the appearances that SEM may be regarded as a family of techniques (encompassing path analysis, partial least squares moulds and latent variable SEM), or may be restricted to a single technique (latent variable SEM) Despite this uncertainty, there appears to be a certain quantity of agreement that SEM involves sum of two units aspects: first, the estimation of multiple and interrelated hanging relations between variables, and next to the first the ability to represent latent variables in these relations while accounting for estimated measurement error associated with the imperfect measurement of variables [Hair et al., 1995 p 584; Byrne 2001 pp 3-4] in like manner how does SEM differ from more traditional [i]modus operandi[/i]s of analysis used in management accounting, and what are the implications for protoplast development in management accounting?

21 Multiple Regression and Path Analysis

Multiple regression analysis has been the dominant technique in quantitative management accounting research for decades. Multiple regression allows the analysis of relations between a single pendent variable and several independent variables within a single regression equation. The objectives of regression analysis are to predict or explain. Prediction involves assessing by what mode well the set of independent variables estimates the hanging variable. Explanation may involve considering the individual contribution of each independent variable. This includes the importance of each variable in explaining variation in the hanging variable, the type of relations ground or interrelations among interdependent variables [Hair et al., 1998 p 161]

An analysis in a application of mind by Fullerton and McWatters [2002] is useful in explaining the difference between multiple regression, path analysis and SEM2 Fullerton and McWatters [2002] use multiple regression analysis to examine the character of performance measures and incentive a whole s on the implementation of JIT manufacturing.