In this dissertation we present Bayesian methods for modeling and analysis of mortgage default data, and develop Bayesian inference procedures to address issues that arise in the analysis. In so doing, we introduce three classes of models for describing default behavior based on aggregate as well as individual level default data. In modeling discrete time aggregate default rates we introduce logistic beta time-series model and consider its extensions such as Markov beta processes and random effects type models to incorporate heterogeneity associated with time. In modeling aggregate number of defaults over time, we present a modulated nonhomogeneous Poisson process model and discuss extensions of the model to deal with heterogeneity associated with different mortgage pools and time intervals. To model individual level default data, we introduce Weibull, generalized Gamma and mixture proportional hazards models and develop Bayesian analysis of these models as well as their heterogeneous extensions. In addition, based on a decision theoretic approach we introduce a loan maintenance framework to illustrate how results from Bayesian mortgage default models can be used for developing preventive maintenance strategies in managing mortgage default risks.