This thesis began with an introduction and literature review in Chapter 1. In Chapter 2, I propose a new intertemporal asset-pricing model based on heterogeneous beliefs to bring together the concurrent theories that could generate value and momentum effects. In this model, I assume that such behaviour occurs simply due to an agnostic view of forecasting returns considering the dominant strategy in the market. Given the endogenous price determination in the model, individuals were expected to adjust their own strategies to match the dominant strategy to obtain higher profits (from more accurate fore- casts). The idea was to bridge the literature on intertemporal asset allocation with the one on heterogeneous beliefs. In Chapters 3 and 4, I consider the empirical problem of implementing Markowitz (1952) mean-variance optimisation on a portfolio of stocks. In particular, I focus on the out-of-sample performance of the minimum-variance portfolio obtained from the use of asset group information and regularisation methods to obtain more stable estimates of the parameters in the model. Specifically, in Chapter 3, I introduce the use of regularisation methods to the portfolio selection problem and a literature review on the subject. In Chapter 4, I propose two alternative approaches for the use of the group structure information and to obtain more stable and regularised minimum-variance portfolios. I show that these procedures produce significantly better results in the portfolios compared with the unconstrained minimum-variance portfolios estimated from the whole data set in terms of portfolio variance and the Sharpe ratio.