17. SEMinR Series | Evaluating Formative Model | Step 1: Redundancy Analysis & Step 2: Collinearity

Evaluating a formative measurement model hinges on two checks that must be planned before data collection: redundancy analysis for convergent validity and variance inflation factors (VIF) for collinearity. Convergent validity in formative models is not assessed through the usual reflective logic. Instead, it measures how strongly a formatively specified construct correlates with an alternative, reflectively measured variable representing the same concept—using a “global single item” as the alternative measure. Researchers must include this global item in the questionnaire at the design stage; once the data are collected, the convergent validity test can’t be performed because the alternative measure is missing.

The redundancy analysis procedure treats each formative construct separately. For a construct like “Vision” measured formatively by multiple indicators, the questionnaire also includes a single global item capturing the essence of Vision (e.g., “I believe in the vision of the organization”). After estimating a simple model linking the formative construct to that global item, the resulting path coefficient is used to judge convergent validity. A commonly cited benchmark is a correlation of at least 0.78 between the formative construct and the global single item; the transcript notes that this threshold implies the construct explains more than 50% of the alternative measure’s variance. In the worked example, the path coefficient from Vision to Vision_Global is 0.818, clearing the recommended threshold (the transcript also references 0.70 as a minimum). Passing this test provides support that the formative construct has convergent validity.

The second step—collinearity—addresses a different risk unique to formative models: indicator weights become unstable when formative indicators are highly correlated. High correlation inflates standard errors of indicator weights and increases the chance of type II errors (false negatives). The standard diagnostic is the variance inflation factor (VIF). VIF values of 5 or above signal collinearity problems, prompting corrective actions such as eliminating or merging indicators, or restructuring the measurement model (for example, using higher-order constructs).

In the practical workflow, results are stored in an R summary object (the transcript references inspecting a “summary_simple” object and then drilling into a validity section containing “VIFore items”). While VIF is also computed for reflective indicators, those values aren’t interpreted here because reflective indicators are expected to correlate; the focus stays on VIF for formative indicators. In the example, all formative-indicator VIF values fall below 5, so collinearity is not a concern.

With redundancy analysis confirming convergent validity and VIF diagnostics showing no collinearity issues, the process is ready to move to the next evaluation step: indicator weights.