@article{2017:hirschmann:multilevel,
title = {Multilevel Structural Equation Modelling in Marketing and Management Research},
year = {2017},
note = {Marketing research often involves two levels (e.g., international
customers nested within countries or cultures). Multilevel modelling
(MLM) allows one to accurately model lower-level (level 1) effects and
the surrounding (level 2) context in addition to various
interrelations between the levels. Hierarchical linear modelling (HLM)
is the most prevalent methodology for MLM. However, HLM does not
handle latent constructs and, compared to multilevel structural
equation modelling (MSEM), has several disadvantages. For example, HLM
does not account for measurement errors. MSEM is increasingly
discussed in methodological papers and textbooks but is seldom
applied. A systematic, nontechnical guide for conducting an MSEM
analysis is lacking, and a systematic review of current and future
applications would benefit future research. Our first research
objective is to introduce scholars to MSEM and its advantages relative
to HLM and to provide novel insights into the topics and flaws of
MSEM/HLM use in extant studies. This manuscript contributes to the
literature by providing a review of 527 marketing and management
studies in which HLM and MSEM are applied. Four categories of
multilevel hypotheses found in studies published in 22 leading
journals over 20 years are discussed and differentiated into five
characteristic research fields (general marketing, international
marketing, international management, general management, and human
resource management). Our second research objective is to address the
requirements, options, and challenges of MSEM. This manuscript
contributes to the literature by systematically discussing the
sampling, measurement, and methodological requirements, options, and
challenges of MSEM-based studies. In doing so, it provides a
nontechnical explanation and a systematic step-by-step procedure for
designing and conducting a cross-sectional MSEM study that tests
hypotheses across levels. Additionally, our empirical study presents
the results for three frequently employed types of MSEM models:
cross-level effects, cross-level interactions, and cross-level effects
and interactions. The latter is presented for the first time in
business research. The manuscript reveals that MSEM is seldom employed
in marketing and management research, although various possible
applications exist. It provides a systematic illustration and
interpretation of MSEM that scholars can consult to conduct future
MSEM-based studies. Finally, promising directions and major challenges
for future MSEM research are examined and are organized as follows:
conceptualization issues (sample size, position of outcome variables,
and control variables), methodological shortcomings (level-specific
preliminary tests and their implementation in software), and general
advice (multi-item scales, computation times, and model convergence).},
journal = {Marketing ZFP},
pages = {50--75},
author = {Hirschmann, Johannes and Swoboda, Bernhard},
volume = {39},
number = {3}
}