Modelling Investment Strategies: Bayesian Learning, Regime Switches and Evolutionary Finance

In this research project we endeavour to model a financial marketplace dominated by a few interacting large institutional investors and draw conclusions about the financial market dynamics that this interaction gives rise to. More specifically, we study the problem of institutional investors, such as pension funds and life assurance companies, which operate in an environment of uncertain cash inflows and uncertain payouts with a minimum threshold.

Investment strategies are taken as model primitives and an artificial financial market is populated by multiple investor types. Trading and investment decisions take place in discrete time. There exist a certain predetermined number of long-lived risky assets paying a random amount of dividends at each discrete point in time, as well as a risk-free asset with a constant interest rate. The
risky assets generate random dividend intensities. Asset payoffs are aggregated and paid to investors at the end of each time period. The general economy is assumed to follow a hidden Markov model with two states, corresponding to normal and recessionary regimes. The existence of a minimum consumption constraint implies the possibility of bankruptcy. The effect of this occurrence on market clearing is modelled explicitly.

We solve our model by means of numerical simulation. The size of the wealth endowment of the different agents is monitored through time over a large number of simulation runs. Our results suggest that both trend following and value investing strategies can be selected by the market under different circumstances. These two results lead
to markedly different outcomes for the economy, as the prevalence of the trend following style leads to destabilization of the marketplace, volatility clustering and severe deflationary spirals. Dividend yield and modified regime-switching CAPM strategies are
never selected by the market.