Learning and strategic asset allocation

This thesis investigates whether or not models that portray the relationship between what an investor learns and how he allocates his portfolio can explain phenomena related to household behaviour in the stock market. Endogenous modelling of household learning is utilised, which builds on a growing literature called bounded rationality with increasing explanatory power, offering an alternative to the classical rational expectations theory. Such phenomena include firstly why households often hold portfolios that are little diversified, secondly why household beliefs about the stock market exhibit widespread heterogeneity despite past data being publicly available and lastly whether or not they employ strategic motives in the stock market and whether they complement others’ actions or substitute them for their own. In particular, Paper 1 addresses the observation that a significant number of investors hold concentrated portfolios, apparently forgoing the benefits of diversification. In a static portfolio choice model with limited capacity constraints, Van Nieuwerburgh & Veldkamp (2010) show that the observed lack of diversification is rational amongst investors with a strong preference for an early resolution of uncertainty. This paper studies whether or not endogenous information acquisition can also rationalise the observed peculiarity within a dynamic portfolio choice model. It is found that in the steady state, a hedging demand component appears in the optimal portfolio, which attempts to spread risk across the investor’s investment horizon: The investor chooses additional information precision the smaller is the compensation for bearing risk. A numerical approximation to the agent’s decision rule suggests that the research question can be answered in the affirmative. This is for an investor who is indifferent to the time of uncertainty resolution and is risk averse over wealth.Paper 2 tackles the question “Is demand for information positively correlated with returns?” Results from this study indicate “no” in any permanent sense. Using a dynamic model of endogenous information acquisition and portfolio choice, simulations reveal that once an agent learns the underlying process behind returns, his demand for information is constant and hence acyclical. This is surprising given survey results found by Coibion et al (2015), which, along with the PATeR survey 2014 wave, document widespread heterogeneity in agent beliefs. Hence it appears that economic agents in reality do not treat returns as though they are driven by an underlying, learnable process. This paper contributes to literature on bounded rationality and limited processing of information. It also finds a rationale for belief heterogeneity: ignoring realised returns and observing signals that support prior beliefs allows degenerate distributions to emerge. This result can be produced under the assumption that the investor does not access public information.Lastly, the third paper addresses whether or not households behave strategically regarding each other when making stock market participation decisions. The study examines the contribution of strategic considerations in stock market return expectations on the demand for risky assets empirically, exploiting novel data from the 2014 PATeR survey wave, representative of the population by age and wealth. The strategy is to identify whether individual stockholding decisions are consistent with strategic substitutes or complements prevailing in the stock market, under the null hypothesis of efficiency. The study finds evidence for strategic complementarity and additional information variables can explain this effect. Given the substantial heterogeneity in expectations and perceptions of returns, and the relatively low degree of sophistication of the median investor identified in the empirical literature, the project concludes that as strategic substitutes prevail even amongst them, a portion of the excess volatility observed in stock markets may be driven by expectational motives in coordination.