In this paper we consider the following interpretation of the formal definition of time consistency: a policy is time consistent if and only if the future planned decisions are actually going to be implemented. In particular for risk averse multistage stochastic programming, we discuss a CVaR based portfolio selection problem and compare a time consistent formulation to a inconsistent one. For the latter, we discuss how different planned and implemented decisions can be and develop a new way of measuring the impact of a inconsistent policy on the objective function. In other words, we argue that the first stage decision of a time inconsistent policy is sub-optimal and we propose a methodology to compute the related sub-optimality gap. For portfolio selection example, we present a sensitivity analysis by computing this gap for different planning horizons and risk aversion levels. Finally, to motivate the use of the proposed time consistent formulation, we develop a suitable economic interpretation for its recursive objective function based on the certainty equivalent of the related preference function.