The consumption characteristics of a textual piece are influenced by both the core-content (i.e., what is being conveyed) and its stylistic attributes (i.e., how it is being conveyed). We present an approach to model stylistic attributes in text and leverage a multi-cause deconfounder model to estimate the causal effect of stylistic attributes towards a target goal. We show that our approach can identify causally significant attributes along with the ones considered important by conventional supervised approaches. Furthermore, we demonstrate using performance comparison on classification tasks that our approach does not compromise on the modeling capabilities. We believe that such a model can be valuable towards providing statistical feedback to an author to improve on certain style attributes to better achieve a target objective.