Search

Schuberth, F., Rosseel, Y., Rönkkö, M., Trinchera, L., Kline, R. B., & Henseler, J. (2023). Structural Parameters under Partial Least Squares and Covariance-Based Structural Equation Modeling: A Comment on Yuan and Deng (2021). Structural Equation Modeling-A Multidisciplinary Journal, 30(3), 339-345. https://doi.org/10.1080/10705511.2022.21 --- Funding: Jorg Henseler gratefully acknowledges financial support from FCT Fundação para a Ciência e a Tecnologia (Portugal), national funding through a research grant from the Information Management Research Center—MagIC/NOVA IMS (UIDB/04152/2020). In their article, Yuan and Deng argue that a structural parameter under partial least squares structural equation modeling (PLS-SEM) is zero if and only if the same structural parameter is zero under covariance-based structural equation modeling (CB-SEM). Yuan and Deng then conclude that statistical tests on individual structural parameters assessing the null hypothesis of no effect can achieve the same purpose in CB-SEM and PLS-SEM. Our response to their article highlights that the relationship they find between PLS-SEM and CB-SEM structural parameters is not universally valid, and that consequently, tests on individual parameters in CB-SEM and PLS-SEM generally do not fulfill the same purpose. publishersversion published

Principal Topic Why some companies grow, and others do not? Answering this question has been one of the main tasks for entrepreneurship researchers (Audretsch et al., 2014; Gilbert et al., 2006; McKelvie & Wiklund, 2010). Differences in managerial motivation is one explanation for varying growth rates in entrepreneurial firms: variation among human motivation and abilities to act has been claimed to have important effects on all phases of the entrepreneurial process (Tominc & Rebernik, 2007), and this relates to the actual firm growth (Delmar & Wiklund, 2008). On a more general level, motivation and the effort that an entrepreneur intends to make have been referred to as growth intentions (Hermans et al., 2015) or growth orientation when examined on a firm level (McKelvie et al., 2017). However, the empirical evidence shows that the magnitude of effects of various motivational constructs ranges from small to medium at best (Levie & Autio, 2013; McKelvie et al., 2017). This fairly weak relationship presents a puzzle: intention is one of the key determinants of human action and owner-managers should have agency over their firms. Based on these considerations, the growth intention of an owner-manager should have a strong impact on firm growth. To explain this disconnection between theory and empirical evidence, researchers have proposed that managers might have limited volitional control over firm growth (Delmar & Wiklund, 2008) or that the relationship between intention and growth is complex and distant and thus not well-captured by a linear, direct effect (Levie & Autio, 2013). In this paper, we develop a two-part argument regarding how growth intention affects growth. First, we argue that the higher growth intentio the better probability of higher growth. This argument is consistent with motivational theories applied in prior research. Second, we also argue that high growth intention increases firm’s probability of failure by lacking growth or even declining over time. The key rationale is that aiming for high growth necessarily increases the risk level (Sapienza et al., 2006) and may lead to sacrificing profitability (Davidsson et al., 2009). By considering these two processes that produce opposite outcomes, we hypothesize that the linear effect of growth intention on expected growth is weak or non-existent at best, but that growth motivation increases the variance of growth outcomes. Thus, growth motivation should be viewed as a necessary but insufficient condition (Dul, 2016) for growth. Altogether, while firms lacking intention to grow are less likely to develop (Levie & Autio, 2013, p. 9), increasing growth intention is insufficient for increasing the growth rate but can also have an opposite effect. Method We test the relationship between growth intention and growth using longitudinal, annually collected survey data from 2,243 small and medium-sized Finnish IT companies (NACE codes 6202 and 6201). We measure growth intention by a self-constructed scale with eight five-point Likert questions, developed from existing literature and piloted with several managers from the study sample. The questions are similar to the ones used by McKelvie et al (2017) and reflect the importance of growth in general, risk resilience for growth achievement and intentions for international expansion. The survey provided us with 4,221 observations collected between 2007-2016. Growth was operationalized as relative change in revenue using one and three-year changes, obtained from Orbis database with almost complete revenue history available for the sample up to 2019. Prior studies, demonstrating a clear linear relationship between growth intention and growth, suffer from two endogeneity problems. First, growth and growth intention are likely to form a feedback loop (Delmar & Wiklund, 2008), that cross-sectional designs do not capture. Second, those studies that apply longitudinal datasets and estimate cross-lagged models, ignore the possibility of unmeasured stable between firm differences, such as the quality of opportunity or initial resource endowment, that are likely to affect growth intention and firm growth causing a spurious correlation; this is a general problem with cross-lagged models (Hamaker et al., 2015). To address these concerns, we start by analyzing our data with cross-lagged random intercept models that simultaneously take in to account both problems explained above (Hamaker et al., 2015; Zyphur et al., 2019). Thereafter we test our variance hypothesis using location-scale models (McNeish, 2020) and by graphical means. To account for missing data in our unbalanced panel, we apply FIML estimation. Results and Implications The preliminary results show that past growth has a positive and statistically significant effect on growth intention and growth intention has a statistically significant and positive effect on future growth, supporting the existence of a feedback loop. Yet, the effects are fairly small and random intercepts for the variables are correlated. Thus, at least partly endogeneity can explain positive correlations found between growth and various operationalizations of growth intention or its dimensions in previous studies. More interestingly, the preliminary analysis shows that while the mean growth rate (expected value) is weakly affected by growth intention, the effect of growth intention on the variance of growth rate, i.e. the prevalence of both extreme positive and negative outcomes is much more pronounced. Additionally to some earlier studies about firm growth being performed with low accuracy (Wright & Stigliani, 2013), there are available research questions concerning firm growth decisions, modes, patterns and reasoning that have received insufficient attention (Clarysse et al., 2011; Iacobucci & Rosa, 2010). This paper contributes to ongoing entrepreneurship research conversations by explaining why relying on growth intention in forecasting growth rates is ineffective solution, which obliterates problems of endogeneity. Our conclusion that increasing willingness or intention to grow a firm only weakly influences on the subsequent growth rate carries value for managers in those companies that intend to grow: they should reconsider the way and the key components of growth to invest in. nonPeerReviewed

The article has the following supplementary material available online. 1. Alternative versions of the tables included in the article using different datasets. 2. Datasets as an excel file, including all manipulations. 3. R code that reproduces all tables and figures included in the article. 4. YouTube playlist contains screencast demonstrations and short video lectures https://youtube.com/playlist?list=PL6tc6IBlZmOWOd0OUIHkQMU3kUz1Vk xY.

Discriminant validity was originally presented as a set of empirical criteria that can be assessed from multitrait-multimethod (MTMM) matrices. Because datasets used by applied researchers rarely lend themselves to MTMM analysis, the need to assess discriminant validity in empirical research has led to the introduction of numerous techniques, some of which have been introduced in an ad hoc manner and without rigorous methodological support. We review various definitions of and techniques for assessing discriminant validity and provide a generalized definition of discriminant validity based on the correlation between two measures after measurement error has been considered. We then review techniques that have been proposed for discriminant validity assessment, demonstrating some problems and equivalencies of these techniques that have gone unnoticed by prior research. After conducting Monte Carlo simulations that compare the techniques, we present techniques called CICFA(sys) and χ2(sys) that applied researchers can use to assess discriminant validity. peerReviewed

Supplemental Material, sj-pdf-1-orm-10.1177_1094428121991907 for Eight Simple Guidelines for Improved Understanding of Transformations and Nonlinear Effects by Mikko Rönkkö, Eero Aalto, Henni Tenhunen and Miguel I. Aguirre-Urreta in Organizational Research Methods

Supplemental Material, Online_supplement_1_-_Chi2(1)_scaling_example for An Updated Guideline for Assessing Discriminant Validity by Mikko Rönkkö and Eunseong Cho in Organizational Research Methods

We examined the sustainability of the KiVa antibullying program in Finland from its nationwide roll‐out in 2009 to 2016. Using latent class analyses, we identified four different patterns of implementation. The persistent schools (43%) maintained a high likelihood of participation throughout the study period. The awakened (14%) had a decreasing trend during the first years, but then increased the likelihood of program participation. The tail‐offs (20%) decreased in the likelihood of participating after the third year, and the drop‐offs (23%) already after the first year. The findings suggest that many schools need support during the initial years to launch and maintain the implementation of evidence‐based programs; yet a large proportion of schools manage to sustain the program implementation for several years. The logistic regression analyses showed that large schools persisted more likely than small schools. Lower initial level of victimization was also related to the sustainability of the program. Finally, persistent program participation was predicted by several school‐level actions during the initial years of implementing the program. These results imply that the sustainability of evidence‐based programs could be enhanced by supporting and guiding schools when setting up the program during the initial implementation. peerReviewed

Supplemental Material, Online_supplement_4_-_Tutorial_on_calculations for An Updated Guideline for Assessing Discriminant Validity by Mikko Rönkkö and Eunseong Cho in Organizational Research Methods

Supplemental Material, Online_supplement_2_-_Full_simulation_code for An Updated Guideline for Assessing Discriminant Validity by Mikko Rönkkö and Eunseong Cho in Organizational Research Methods

Supplemental Material, Online_supplement_5_-_Deriving_a_cutoff_based_on_simulation_results for An Updated Guideline for Assessing Discriminant Validity by Mikko Rönkkö and Eunseong Cho in Organizational Research Methods

Templeton and Burney (2017) proposed a two-step normality transformation as a remedy for non-normally distributed data, which are commonly found in AIS research. We argue that, rather than transforming the data toward normality, researchers should first seek to analyze and understand the sources of non-normality. Using simulated datasets, we demonstrate three sources of non-normality and their consequences for regression estimation. We then demonstrate that the two-step transformation cannot solve any of these problems and that each source of non-normality can be handled with alternative, existing techniques. We further present two empirical examples to demonstrate these issues with real datasets. peerReviewed