In this paper, we study the problem of modelling the dependence of defaults in several sectors. We consider a network-based model for the default data sequences and model the sequences by a Markov chain model. The new model provides a flexible paradigm for portfolio credit risk assessment. We evaluate two important risk measures, namely, crisis value-at-risk (CRVaR) and crisis expected shortfall (CRES). Numerical experiments are given to illustrate the practical implementation of the model. We also perform empirical studies of the model using real default data and analyse the empirical behaviours of the risk measures arising from the model.