Predictive Property AnalyticsMIAC’s indices provide a robust picture of historical and current house prices. However, much effort is also directed at understanding how these time series may evolve in the future. There are many important reasons for undertaking these investigations such as
- Stress testing
- IFRS 9
- Portfolio valuation
- Seasonality. House prices generally are their lowest in spring and highest in summer.
- Trending. House prices generally increase at a fairly constant rate over time.
- Autocorrelation. If house prices decline one month, they are more likely to decline in the next month.
Whilst any house price predictions would have to take these features into account, they are insufficient in isolation to produce reliable estimates of future movements. This is why practitioners look to other factors that impact house prices. There are many factors that impact house prices and these can be used via machine learning and other methods to gauge future movements.
- Interest rates
- Economic growth (GDP)
- Mortgage availability
- Supply and demand
- Government policy
Single Path Predictions
The Office for Budget Responsibility provides five-year forecasts for the economy and public finances twice a year. Included within these forecasts is one for UK house prices. The four most recent projections are shown in Figure 1. The historical price used is the MIAC UK ‘All’ index.
The reliability of any prediction model decreases with time and therefore the further out in time you want to predict the more error is going to be introduced. Long terms contracts held on the banking book such as mortgages and bonds require house price scenarios that extend many years and so the usual empirical approach to modelling specific house price scenarios quickly becomes inaccurate. One of the only things you can say about any specific long term house price predictions is that they will be wrong.
MIAC’s approach is one of prediction of distributions rather than single specific pathways. Using distributions allows us to assign probabilities to any given scenario and therefore provide a much more holistic view of the future.
Figure 3 and Figure 4 illustrates the difference between using a single path approach to the whole distribution when undertaking predictive analytics.
This is also especially useful when determining the likelihood of any given single scenario. For example the Bank of England adverse scenario appears to be on the extreme tail of the distribution indicating that it is an outlier as intended and specific probabilities can be assigned to this pathway at any given point in time.
Get in touch today to find out how MIAC can help you use predictive property analytics for risk management and many other uses.