The research project has two broad objectives and provides novel results in the literature of structural change or change-point tests. The first objective is to provide two new methods for restoring the non-monotone power problem of a large family of structural breaks tests that have been widely used in econometrics and statistics, as well as to show that these methods have additional contributions and can be extended to: (i) tests for a change in persistence, (ii) partial sums tests of cointegration and (iii) tests for changes in dynamic volatility models. The significance of these methods is demonstrated via the consistency of the long-run variance estimator which scales the change-point statistics, the asymptotic properties of the tests, their finite sample performance and their relevance in empirical applications and policy analysis. The second objective is threefold: First, to show that ignoring structural changes in financial time series yields biased and inconsistent risk management (Value at Risk, VaR and Excess Shortfall, ES) estimates and consequently leads to investment misallocations. Second, to propose methods for evaluating the stability of financial time series sequentially or on-line which can be used as a quality control procedure for financial risk management as well as to show that monitoring implied volatilities yields early warning indicators of a changing risk structure. Moreover we show that model averaging in the presence of structural breaks as well as other model uncertainties involved in risk management estimates, can provide robust estimates of VaR and ES. New results are derived on the optimal weights for model averaging in the context of dynamic volatility models and asymmetric loss functions. Third, we propose a novel way to construct prediction-based change-point statistics that reduce the detection delay of existing sequential tests and provide a probability about the likelihood of a structural change.