Why PLS-SEM for Technology Adoption Research?
A brief explanation of why Partial Least Squares Structural Equation Modeling (PLS-SEM) is the appropriate statistical method for testing technology adoption frameworks like UTAUT2.
The Core Question
When studying what predicts whether people will adopt a new technology or organizational procedure, we need to test multiple factors simultaneously. This is where the choice of statistical method becomes critical.
Two Different Approaches, Two Different Questions
Correlation Analysis (Pearson Correlation)
Question it answers: “Do these two variables move together?”
What it does: Tests the relationship between two variables at a time, ignoring all other factors.
Type of relationship: Bivariate (two variables only)
Structural Equation Modeling (PLS-SEM)
Question it answers: “What uniquely predicts the outcome when all factors compete simultaneously?”
What it does: Tests multiple predictors together in a complete theoretical framework, controlling for all other variables.
Type of relationship: Multivariate (multiple variables simultaneously)
A Hypothetical Example
Scenario: Predicting Cloud Storage Adoption
Imagine we’re studying what predicts whether IT professionals will adopt a new cloud storage system. We’re testing these potential predictors:
- Performance Expectancy: “Will it improve my work efficiency?”
- Effort Expectancy: “Is it easy to use?”
- Social Influence: “Do my colleagues use it?”
- Facilitating Conditions: “Do I have the resources and support to use it?”
Using Correlation Analysis:
We might find that ALL four factors correlate with adoption intentions:
- Performance Expectancy: r = 0.52
- Effort Expectancy: r = 0.45
- Social Influence: r = 0.38
- Facilitating Conditions: r = 0.54
Conclusion from correlations: “All four factors are related to adoption intentions!”
Using PLS-SEM:
When we test all four factors together in a structural model, we might find:
- Performance Expectancy: β = 0.28, p = 0.003 (Significant)
- Effort Expectancy: β = 0.05, p = 0.582 (Not significant)
- Social Influence: β = 0.11, p = 0.234 (Not significant)
- Facilitating Conditions: β = 0.31, p = 0.001 (Significant)
Conclusion from PLS-SEM: “Only Performance Expectancy and Facilitating Conditions uniquely predict adoption intentions when all factors compete together.”
Why the Difference?
Effort Expectancy correlates with adoption intentions on its own, but when Performance Expectancy and Facilitating Conditions are in the model, Effort Expectancy doesn’t add unique predictive power. The other constructs are “absorbing” its contribution.
This is the power of structural modeling – it shows what matters above and beyond the other predictors.
Why This Matters for Theory Testing
Technology adoption frameworks like UTAUT2 propose a system of relationships – not just individual correlations. These frameworks suggest that multiple factors work together to predict behavior. To properly test such a framework, we need a method that:
- Tests all relationships simultaneously
- Controls for competing explanations
- Shows which factors have unique predictive power
- Reveals the overall explanatory power of the model
PLS-SEM does all of this. Correlation analysis does not.
Comparison Table
| Aspect | Correlation Analysis | PLS-SEM |
|---|---|---|
| Number of variables tested | Two at a time | Multiple simultaneously |
| Controls for other factors | No | Yes |
| Tests theoretical frameworks | No | Yes |
| Shows unique contribution | No | Yes |
| Calculates variance explained (R²) | Only for two variables | For entire model |
| Appropriate for complex models | No | Yes |
| Standard for UTAUT/UTAUT2 | No | Yes |
The Role of Correlation in PLS-SEM Studies
Correlation analysis isn’t useless in PLS-SEM research – it serves an important but different purpose:
Correlation’s Role: Data Quality Check
In PLS-SEM studies, correlations among variables are examined to check for multicollinearity – when predictors are so highly correlated (typically > 0.9) that they cause statistical problems.
Purpose: Quality assurance during data screening (Stage 1)
NOT for: Testing hypotheses or reporting findings
The Three-Stage PLS-SEM Process
According to Hair et al. (2014), PLS-SEM follows a systematic three-stage approach:
Stage 1: Data Screening
- Check for missing data and outliers
- Test multicollinearity (using correlations and VIF)
- Verify data meets analysis assumptions
Stage 2: Measurement Model Assessment
- Validate that survey questions measure intended constructs
- Check reliability and validity
- Ensure constructs are distinct from each other
Stage 3: Structural Model Assessment
- Test hypothesized relationships (path coefficients)
- Calculate variance explained (R²)
- Determine which hypotheses are supported
- This is where research findings come from
When to Use Which Method
Use Correlation Analysis When:
- You want to explore general associations between two variables
- You’re doing preliminary data screening
- You’re checking assumptions for other analyses
- Your research question is specifically about bivariate relationships
Use PLS-SEM When:
- Testing a theoretical framework with multiple constructs
- Your model has multiple predictors of an outcome
- You need to know unique contributions of each predictor
- You’re doing explanatory or predictive research
- Your research uses frameworks like UTAUT, UTAUT2, TAM, TPB
Why Standard Methods Matter
Technology adoption research using UTAUT2 has established methodological standards. Using PLS-SEM for such studies:
- Follows accepted practices in the field
- Makes results comparable to other studies
- Properly tests the theoretical framework as intended
- Meets expectations of reviewers and committee members
- Provides the most appropriate statistical evidence
Bottom Line
PLS-SEM isn’t “better” than correlation in all contexts – they answer different questions. For testing theoretical frameworks like UTAUT2 with multiple simultaneous predictors, PLS-SEM is the appropriate choice because it tests the system of relationships that the theory proposes, not just individual associations.
Additional Resources
For those interested in learning more:
- Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2014). A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). SAGE Publications.
- Hair, J. F., Ringle, C. M., & Sarstedt, M. (2011). PLS-SEM: Indeed a Silver Bullet. Journal of Marketing Theory and Practice, 19(2), 139-152.
A Primer on Partial Least Squares Structural Equation Modeling
Recent Advances and Applications in Partial Least Squares Structural Equation Modeling
PLS-SEM for Building and Testing Behavioral Causal Theory