Pathway analysis allows joint consideration of multiple SNPs belonging to multiple genes, which in turn belong to a biologically defined pathway. We develop a Bayesian hierarchical model by fully modeling the three-level hierarchy, namely, SNP-gene-pathway that is naturally inherent in the pathway structure, unlike the currently used ad hoc ways of combining such information. To handle the high dimensionality, we regularize the effects at each level through appropriate choice of hierarchical priors. A key advantage of the joint modeling is that not only can we find associated pathways but also the associated genes within the significant pathways, and the associated SNPs within the significant genes. Such a formal mechanism for testing of components of a significant pathway is useful for follow-up studies. We use Hierarchical False Discovery Rate for multiplicity adjustment of the entire inference procedure. We conduct simulations with samples generated under realistic linkage disequilibrium patterns. We find that our method has higher power than some current approaches for identifying pathways with multiple modest-sized variants. The method is illustrated on real cancer data.