As commitment levels to pro-environmental activities are usually coded as ordered categorical variables, we argue that the multivariate ordered probit model is an appropriate tool to account for the effect of common observable and unobservable variables on joint pro-environmental behaviors. However, exploring in-depth joint pro-environmental behaviors using the multivariate ordered probit model requires not only to assess whether some variables are found to be significant, but also to calculate joint probabilities, conditional probabilities and partial effects on these quantities. As an illustration, we explore the joint commitment levels of households to recycling of materials in France. We show that beyond the estimation of the multivariate ordered probit model, much can be learned from the calculation of the aforementioned quantities. (JEL C35, Q53)