Structural equation modeling applications in agricultural technology adoption and extension research

Introduction

Take a farmer's decision to adopt drip irrigation. It depends on how much the farmer knows about the technology, whether neighbors have adopted it, the cost relative to expected returns, trust in the extension source recommending it, and perhaps a dozen other factors that influence each other in ways simple correlation cannot capture [1] . Structural equation modeling was designed precisely for this kind of complexity it handles latent variables, simultaneous equations, and measurement error in a single analytical framework [2] . SEM entered agricultural research relatively late compared to psychology and marketing, where it had been standard since the 1980s [3] . Early applications in extension focused mainly on path analysis with observed variables, but the shift toward full latent variable models accelerated after 2010 as software became more accessible and training materials more widely available [4] . By 2020, SEM had appeared in journals spanning agricultural economics, rural sociology, and extension education. Yet popularity doesn't guarantee quality. Methodological reviews in psychology and management have consistently found that many published SEM applications fall short of recommended practices insufficient sample sizes, poor model fit reporting, and failure to test alternative specifications [5] . Whether the same problems affect agricultural adoption research hasn't been systematically assessed. This research aimed to map SEM usage patterns in extension research, evaluate methodological quality against established benchmarks, and identify gaps where the field needs improvement.

Conceptual Model

The review was structured around Kline's ten commandments of SEM practice [2] , which provide a consensus framework for evaluating methodological quality. These include: adequate sample size (minimum 10:1 ratio of cases to parameters), correct model identification, use of multiple fit indices, testing of measurement models before structural models, reporting of standardized and unstandardized solutions, evaluation of modification indices, examination of alternative models, assessment of www.extensionjournal.com measurement invariance for group comparisons, transparent reporting of effect sizes, and acknowledgment of limitations. These ten criteria were expanded into a 42-item checklist covering specification, estimation, evaluation, and reporting dimensions.

Material and Methods

Material

A systematic review protocol following PRISMA guidelines was registered prior to data collection. Five databases were searched: Web of Science, Scopus, CAB Abstracts, AGRIS, and Google Scholar. Search terms combined SEM-related terms ("structural equation," "path analysis," "latent variable modeling") with agriculture and extension terms. Inclusion criteria required empirical application of SEM to agricultural technology adoption or extension servicerelated outcomes, publication in peer-reviewed journals between January 2005 and December 2023, and English or Spanish language. The review was anchored at the National University of Asuncion, Faculty of Agricultural Sciences, Asuncion, Paraguay.

Methods

Two reviewers independently screened 684 initial records, with disagreements resolved through discussion with a third reviewer. The final sample comprised 118 articles. Each article was coded using the 42-item methodological quality checklist, with items scored as present (1), partially present (0.5), or absent (0). Inter-rater reliability was assessed using Cohen's kappa (κ=0.83). Articles were categorized by publication period (pre-2015 vs. post-2015), theoretical framework used, and geographic region. Descriptive statistics summarized quality dimensions. Chi-square tests compared quality indicators between periods. All analyses used SPSS version 27.

Pilot Research

The 42-item checklist was pilot-tested on 15 randomly selected articles from the initial pool. The pilot revealed that four items were ambiguous in distinguishing between partial and full compliance, leading to revised scoring rubrics. Two items on modification index reporting were merged because reviewers could not reliably separate them. The pilot also established that coding each article required approximately 45 minutes, allowing realistic timeline planning for the full review. Inter-rater agreement during the pilot was κ=0.78, which improved to κ=0.83 after rubric revisions. The radar plot shows improvement across all seven quality dimensions, but the shape reveals persistent weaknesses.

Results

Invariance testing and alternative model examination remain the smallest spokes even in the post-2015 period. Measurement validity improved the most dramatically (from 51.2% to 74.6%), likely reflecting greater awareness of confirmatory factor analysis as a prerequisite step.

Discussion

The fivefold increase in SEM usage since 2005 reflects a field grappling with the limitations of simpler methods for modeling adoption a process that involves latent constructs, indirect effects, and measurement error that regression cannot accommodate [1] . The improvement in quality indicators is encouraging. But two persistent gaps stand out. First, fewer than half of recent articles tested measurement invariance. Without invariance testing, comparisons across farmer groups (by region, gender, or education level) are questionable because the measurement instrument may not function equivalently across subpopulations [6] . This is particularly problematic in extension research where subgroup comparisons are common and policy-relevant. Second, the low rate of alternative model testing means most authors presented their initial specification as the final answer without exploring whether competing models fit the data equally well [7] . This confirmation bias undermines the credibility of causal claims. The Theory of Planned Behavior's strong model fit is consistent with its track record in adoption research [8] , but the relatively poor performance of Self-Determination Theory suggests that newer theoretical imports from psychology may need adaptation before they translate cleanly into agricultural contexts. Researchers applying unfamiliar frameworks should invest in extensive pilot work to refine their measurement models before committing to full SEM estimation [9] .

Limitations

The review was restricted to English and Spanish language publications, potentially missing SEM applications published in other languages.

References

1. Feder G, Just RE, Zilberman D. Adoption of agricultural innovations in developing countries: a survey. Economic Development and Cultural Change. 1985;33(2):255-298.
2. Kline RB. Principles and practice of structural equation modeling. Guilford Press. 2015;4th ed:1-534.
3. Bagozzi RP, Yi Y. On the evaluation of structural equation models. Journal of the Academy of Marketing Science. 1988;16(1):74-94.
4. Teo T. A comparison of non-nested models in explaining technology adoption. British Journal of Educational Technology. 2012;43(1):E6-E10.
5. Shah R, Goldstein SM. Use of structural equation modeling in operations management research. Journal of Operations Management. 2006;24(2):148-169.
6. Chen FF. Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling. 2007;14(3):464-504.
7. Tomarken AJ, Waller NG. Structural equation modeling: strengths, limitations, and misconceptions. Annual Review of Clinical Psychology. 2005;1:31-65.
8. Wauters E, Bielders C, Poesen J, Govers G, Mathijs E. Adoption of soil conservation practices in Belgium. Land Use Policy. 2010;27(4):1082-1089.
9. Anderson JC, Gerbing DW. Structural equation modeling in practice: a review and recommended twostep approach. Psychological Bulletin. 1988;103(3):411-423.
10. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis. Structural Equation Modeling. 1999;6(1):1-55.
11. Hair JF, Black WC, Babin BJ, Anderson RE. Multivariate data analysis. Pearson. 2019;8th ed:1-813.
12. Davis FD. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly. 1989;13(3):319-340.
13. Ajzen I. The theory of planned behavior. Organizational Behavior and Human Decision Processes. 1991;50(2):179-211.
14. Rogers EM. Diffusion of innovations. Free Press. 2003;5th ed:1-576.
15. Venkatesh V, Morris MG, Davis GB, Davis FD. User acceptance of information technology: toward a unified view. MIS Quarterly. 2003;27(3):425-478.
16. Byrne BM. Structural equation modeling with AMOS: basic concepts, applications, and programming. Routledge. 2016;3rd ed:1-460.
17. MacCallum RC, Austin JT. Applications of structural equation modeling in psychological research. Annual Review of Psychology. 2000;51:201-226.
18. Kaplan D. Structural equation modeling: foundations and extensions. SAGE Publications. 2009;2nd ed:1276.