Making Models of the Economy
Making models of the economy is hard. There are two reasons for that. First of all, unlike physics for example, the economy cannot be described by a small set of mathematical equations. Instead there are hundreds and hundreds of factors that has to be considered, even for a simple model of the economy. This includes macro factors, such as interest rates, balance of trade figures and unemployment rates, political factors, such as taxation, import regulation and social security policies. Other factors include ever changing psychological biases, technological inventions and geographical characteristics. All of these factors are interrelated, and the relationships vary between businesses and individuals on a micro scale. E.g. certain businesses are very sensitive to currency rates, and certain individuals may change their consumption pattern entirely because of a tax cut, whereas others don’t care.
The second reason for the unpredictability of the economy is known as the butterfly effect. Small changes tend to make a big difference in the long term. The IBM decision to let Microsoft keep the rights to MS-DOS made it possible for a small software company to grow into an empire. Mathematically, this is known as sensitive dependence on initial conditions. Yet, not all decisions change the world, and certainly many aspects of the economy are predictable, even in the long run.
The only way to manage these difficulties is to make statistical models. In a statistical model, the outcome is not given beforehand, but different outcomes have different probabilities to occur. These kinds of models are often used to describe the stock market, using a random walk model. Although such a model normally cannot be used to forecast future stock prices, it can be used to price options, to estimate risks or to select optimal portfolios. Similar models can be used to model any kind of economical time series. Some types of time series are more random, whereas others are smoother and more predictable.
The more you know about the fundamentals of a given time series, the better models you are likely to make. A foreign currency analyst might construct a model for a currency that takes into account interest rates, balance of trade figures, GDP growth rates, psychological factors and financial monetary flows. Such a sophisticated model would probably be much more likely to have predictive power than a simple random walk model. The only problem is that you would also need models of interest rates, export and import, GDP growth etc. Each of these models might require additional factors, most of which would be outside of the expertise of the foreign currency analyst.
Part of the solution to the foreign currency analysts dilemma is the Monte Carlo Modeling Engine. A model created for the Monte Carlo Modeling Engine is automatically able to use the output of other models. This allows the foreign currency analyst to concentrate on making good models of currency exchange rates, and using models from other experts as the input he needs. He may allow other people to use his model as well, perhaps by posting it for download here, at economymodels.
As implied by the name, the Monte Carlo Modeling Engine uses the Monte Carlo method to run the statistical models. This means that the models are run many times, perhaps tens of thousands or millions of times, each time with a different random outcome. These different outcomes can then be analyzed using statistical tools. The result is not a prediction of the future, but an assessment of which outcomes are more or less likely.
External links
Send this page to a friend
Send us your comments