#SmartPLS4 Series 19 - How to Report Higher Order Reflective-Formative Construct?

TL;DR

Internal Marketing is modeled as a higher-order reflective–formative construct, with reflective lower-order components (Vision Development and Rewards) forming a formative higher-order construct.

Briefing Cornell Notes

Briefing

A validated higher-order construct can be reported in SmartPLS 4 only after it clears a specific checklist: redundancy analysis for convergent validity, variance inflation factor checks for collinearity, significance testing of outer weights, and confirmation that outer loadings meet minimum thresholds. In this session, the focus lands on “Internal Marketing” as a higher-order reflective–formative construct—reflective at the lower level, formative at the higher level—built from three reflective lower-order components: Vision Development and Rewards.

Reporting begins after the lower-order constructs have already been validated. With that foundation in place, the higher-order formative construct is assessed using the same core measurement-model requirements: reliability and validity metrics such as Cronbach’s alpha, composite reliability (CR), average variance extracted (AVE), and heterotrait–monotrait ratio (HTMT). Once those lower-level results are accepted, the higher-order formative reporting follows a structured sequence tailored to reflective–formative models.

Step one is redundancy analysis, which is used to establish convergent validity for the formative higher-order construct. Convergent validity here means the formative construct correlates strongly with an alternative reflectively measured global indicator. The method requires planning ahead in the questionnaire: researchers must include a global single-item measure capturing the essence of the formative construct. In this case, the global item asked whether respondents were happy with the vision development and rewards. After data collection, the correlation between the formative construct and the global reflective item should exceed 0.708; that threshold corresponds to the construct explaining more than half of the variance in the alternative measure. The reported correlation met the criterion (0.8), so convergent validity was treated as satisfied.

Step two checks multicollinearity using the variance inflation factor (VIF). If VIF values exceed 5, collinearity becomes a concern and can distort formative weight estimates. The results showed no VIF values above 5 for the higher-order construct, indicating no collinearity issues.

Step three evaluates the significance of the formative indicators through outer weights. This requires bootstrapping to generate t statistics and p values. The indicators—Vision Development and Rewards—showed significant outer weights, with bootstrapping outputs used to report t statistics, p values, and the direction of effects.

Finally, outer loadings are reviewed to ensure indicators are meaningfully related to their constructs. The session notes that outer loadings were above 0.50 and significant for each indicator. With redundancy analysis, VIF, outer-weight significance, and outer-loading thresholds all satisfied, “Internal Marketing” is concluded to be a valid higher-order reflective–formative construct—and the reporting steps in SmartPLS 4 are presented as the practical path to documenting those results.

Cornell Notes

Internal Marketing is treated as a higher-order reflective–formative construct in SmartPLS 4, built from three reflective lower-order components (Vision Development and Rewards). Reporting it requires more than the lower-order validations: the higher-order formative construct must pass redundancy analysis for convergent validity, VIF checks for collinearity, bootstrapped significance tests for outer weights, and confirmation that outer loadings exceed a minimum threshold. Convergent validity uses a global single-item alternative measure; the correlation should be above 0.708, and the session reports 0.8. VIF values stayed below 5, and bootstrapping showed significant outer weights and outer loadings above 0.50. Passing these criteria supports the conclusion that Internal Marketing is a valid higher-order reflective–formative construct.

Why does redundancy analysis matter when a higher-order construct is formative at the top level?

Redundancy analysis is used to establish convergent validity for the formative higher-order construct by checking whether it correlates with an alternative reflectively measured global indicator. The method requires a global single-item measure planned in advance in the questionnaire. In this case, the global item asked whether respondents were happy with the vision development and rewards. The correlation between the formative construct score and that global reflective item should exceed 0.708; the reported value was 0.8, indicating the formative construct explains more than half of the alternative measure’s variance.

What threshold signals a convergent validity problem in this setup, and what value was reported?

The key threshold is 0.708 for the correlation between the formative higher-order construct and the alternative reflectively measured global item. Values above 0.708 indicate no convergent validity issue. The session reports a correlation of 0.8 for Internal Marketing, which clears the threshold.

How does the variance inflation factor (VIF) test protect formative indicators from collinearity issues?

After redundancy analysis, the model checks VIF to detect multicollinearity among formative indicators. If VIF values exceed 5, collinearity can bias formative weight estimates. For the higher-order Internal Marketing construct, the reported VIF values did not exceed 5, so collinearity was not considered a problem.

What must be reported for formative higher-order indicators, and how are significance results obtained?

Formative higher-order indicators are evaluated through outer weights. Significance requires bootstrapping, which produces t statistics and p values for each indicator’s outer weight. The session instructs reporting these statistics (t statistics and p values) for Vision Development and Rewards, and notes that the outer weights were significant after bootstrapping.

Why check outer loadings after validating outer weights in a reflective–formative higher-order model?

Outer loadings confirm that indicators have sufficiently strong relationships with their constructs. The session uses a practical threshold: outer loadings should be greater than 0.50 and significant. For Internal Marketing, outer loadings for each indicator were above 0.50 and significant, supporting the validity conclusion.

Review Questions

What global single-item measure is required for redundancy analysis in a higher-order formative construct, and what correlation threshold indicates success?
Which SmartPLS 4 outputs are needed to report formative higher-order indicator validity (outer weights, t statistics, p values, outer loadings), and why is bootstrapping required?
How do VIF results influence confidence in formative indicator estimates, and what VIF cutoff is used?

Key Points

1
Internal Marketing is modeled as a higher-order reflective–formative construct, with reflective lower-order components (Vision Development and Rewards) forming a formative higher-order construct.
2
Reporting higher-order reflective–formative constructs in SmartPLS 4 requires validating lower-order constructs first, then running higher-order checks.
3
Redundancy analysis for convergent validity uses a planned global single-item alternative measure; the correlation should exceed 0.708 (0.8 was reported).
4
Collinearity among formative indicators is assessed with VIF; values above 5 signal a problem (no VIF issues were reported).
5
Outer weights for formative indicators must be tested for significance using bootstrapping, with t statistics and p values reported.
6
Outer loadings should exceed 0.50 and be significant for each indicator to support the validity conclusion.
7
When redundancy analysis, VIF, outer-weight significance, and outer-loading thresholds all pass, the higher-order reflective–formative construct can be reported as valid.

Highlights

Convergent validity for a formative higher-order construct hinges on redundancy analysis against a global single-item alternative measure; the session uses a 0.708 correlation threshold.

The Internal Marketing model clears the convergent validity bar with a reported correlation of 0.8 and shows no VIF values above 5.

Bootstrapping is essential for reporting formative indicator outer weights, including t statistics and p values for Vision Development and Rewards.

Outer loadings above 0.50 (and significant) provide an additional validity check beyond outer-weight significance.

Topics

Higher-Order Constructs
Reflective–Formative Modeling
SmartPLS 4 Reporting
Redundancy Analysis
Bootstrapping

Mentioned

SmartPLS 4
CR
AVE
HTMT
VIF