Multivariate Model

The multivariate model is a popular statistical tool that uses multiple variables to forecast possible outcomes. Research analysts use multivariate models to forecast investment outcomes in different scenarios in order to understand the exposure that a portfolio has to particular risks. This allows portfolio managers to better mitigate the risks identified through the multivariate modeling analysis. The Monte Carlo simulation is a widely used multivariate model that creates a probability distribution that helps define a range of possible investment outcomes. Multivariate models are used in many fields of finance.

Multivariate models assist with decision making by allowing the user to test out the different scenarios and their probable impact. For example, a particular investment can be run through scenario analysis in a multivariate model to see how it will impact the whole portfolio return in different market situations, such as a period of high inflation or low interest rates. This same approach can be used to evaluate a company’s likely performance, value stock options and even evaluate new product ideas. As firm data points are added to the model, such as same store sales data being released prior to earnings, the confidence in the model and its predicted ranges increase.

Insurance companies are users of multivariate models. The pricing of an insurance policy is based on the probability of having to pay out a claim. Given only a few data points, such as the age of the applicant and the home address, insurers can add that into a multivariate model that pulls from additional databases that can narrow in on the appropriate policy pricing strategy. The model itself will be populated with confirmed data points (age, sex, current health status, other policies owned, etc.) and refined variables (average regional income, average regional lifespan, etc.) to assign predicted outcomes that will be used to price the policy.

The advantage of multivariate modeling is that it provides more detailed “what if” scenarios for decision makers to consider. For example, investment A is likely to have a future price within this range given these variables. As more solid data is put into the model, the predictive range gets tighter and confidence in the predictions grow. However, as with any model, the data coming out is only as good as the data going in. There is also a risk of black swan events rendering the model meaningless even if the data sets and variables being used are good. This is, of course, why the models themselves aren’t put in charge of trading. The predictions of multivariate models are simply another source of information for the ultimate decision makers to think about.