To interpret the results, I would look at all the direct effects (path/regression slope coefficients) first, including X --> M and M --> Y. I would assume that the direct effect X (procrastination) --> M (anxiety) should be positive (the higher the level of procrastination, the higher the anxiety level). I would further assume that the path M (anxiety) --> Y (academic performance) is negative (the higher the anxiety, the lower the performance). This would explain the negative indirect effect (since the indirect effect is equal to the product of the two direct effects X --> M and M --> Y).

The negative direct effect X --> Y appears to make sense (the higher the level of procrastination, the lower the performance).

Hi,

to add a thought. Doesn't it make more sense that anxiety causes procrastination? In addition both could form a feedback loop in which case, your model would not be able to identify both or a single effect. Don't you have any instrumental variables to strengthen the model? A simple partial mediation model is not a strong support for the mediation process.

Best,

Holger

It is a credible point. But in this study, I have considered only these three variables.