SEM - insignificant direct path vs. significant indirect paths

[Janne Brouer]

Dear Forum,

I am conducting SEM analysis in Stata and I have specified a well-fitting measurement model via CFA using the Satorra-Bentler adjustment. Overall, I have good model-fit

• Goodness of fit indices: with my re-specified model, I achieved overall good model fit, e.g. CFI > 0.92, RMSEA < 0.7

When I do my structural model, I find a lot of insignificant paths that force me to reject my hypotheses from my conceptual framework. This stands in great contrast to the empirical studies that I have reviewed where the construct: Environmental Concern (EC) always had a significant association with Behavioral Intention (BI), as well as the original constructs from the Theory of Planned Behavior (Attitudes toward behavior, Subjective Norms, and Perceived Behavioral Control).

Issue:

• I have two behaviors that I am measuring
  • for BI1, the path between EC and BI1is significant (as theory suggests)
  • for BI2, I have a strong insignificant path between EC and BI2, where BI2 is the purchase of second-hand furniture.
    • However, all the indirect paths - from EC to ATB, SN, and PBC are all significant.
    • When I delete the insignificant paths, then the direct path between EC and BI2 is suddenly significant - how can that be?

Thank you very much in advance for your help. For reference, find a picture of my structural model below.

Best,
Janne.