# When can I assume that a variable is exogenous?

I'm looking for the effect of nutritional coffee breaks on MBA performance, and I found that the change to a healthier catering increased the performance of MBA students. Additionally, I want to see if **more exposure to nutritional meals** affects more or less these students' test scores. To do that, I want to use **attendance to measure the intensity of the treatment** (if you attended more, you were more exposed to the treatment). Still, given that the healthier snacks might affect attendance, I first obtain the effect of healthier on attendance, estimating a precise zero. Consequently, can I use attendance as a measure of intensity? or even if I find that healthier coffee breaks do not affect attendance, I should assume that it is endogenous. If the aforementioned answer is correct, can you explain to me why?
Thank you very much.

$$Performance_i = \beta_0 +\beta_1 Attend_i +\varepsilon_i$$
It may be that healthier snacks do not affect attendance, but I would be very concerned making an assumption of the form $$Cov(Attend_i, \varepsilon_i)=0$$. Students who are more likely to perform well are also probably more likely to attend. i.e., there could be an omitted variable, $$Motivation_i$$ such that $$\varepsilon_i =\beta_2 Motivation_i +\eta_i$$.
Then $$Motivation_i$$ is probably correlated with both attendance and performance.