Tuesday, January 22, 2019

The Model Thinker #4 : Modelling Human Actors

Image result for vitrivius man

The first serious chapter on modelling covers human actors.

There are two general approaches to modelling a human actor.

a) The Rational Actor Model

Modelling human actors as being rational has come under a lot of criticism but no other model has the kind of consistency and elegance offered by this approach. By making an assumption that a person will act rationally, you can also come up with behaviour that approximates that of a highly intelligent player.

In the Rational Actor model, individuals choose the action that maximises their payoff. This pre-supposes that a payoff function is defined over a set of possible actions. If an investor is rational, between an investment that generates 10% and 12% but with the same semi-variance, the investor will choose the latter strategy.

Upon further inspection, you will realise that most human beings are hardly rational. One component of rationality is that ability to consider all options. All alternatives must have a preference ordering. Another aspect of rationality is transitiveness. If you like A over B, and B over C. You must like A over C. ( I am hardly rational when it comes to movies )

If you come into the financial markets assuming that all market actors are rational, you will be in serious trouble. Even I have serious quirk which I am not really willing to show the world even though Dr Wealth colleagues have asked me to give more details on my portfolio - one of the quirks I am less proud off is that I like to invest in denominations of eight... For example, I have 8800 shares of UMS.

To blunt the lofty assumptions of Rational Actor theory, economists now add theories of loss aversion and hyperbolic discounting into whatever models they have. We take larger risks when faced with losses, and we apply a stronger discount rate of our immediate future.

b) Rule-based model

Rule based models do not use a payoff function but assumes that people will behave based on specific behaviours. One possible way to model the behaviour of an average Singaporean is that, based on government  data, he is likely to set aside 20% of his take home pay.

The rules-based model may be fixed, but common sense tells us that people adapt to policy changes and changes. When the salary of a person goes up, he might be able to save a larger proportion of his income. He may have to forego savings when he gets retrenched.

The spirit of modelling is to adopt the approach that all models are wrong one way or another. One criticism of modelling the behaviour of human beings is that changes in policy or the environment will produces behavioural responses by those affected. This renders past behaviour inaccurate.

But we know this problem well in quantitative investing.

This is the problem of non-stationarity.

No comments:

Post a Comment