DIY House Price Forecasting

Las Vegas, Nevada, USA, 27 July 2009. Predicting markets is something that investors have been trying to do for centuries, with limited success. What if you had the key to real estate forecasting? There are a few models which may hold the secret.

The housing crisis in the US is largely the cause of the global economic slump, and many believe when it comes back, both in the US and globally, we will see a return to growth and prosperity. So it is fitting that we examine how to forecast home prices. [br]

 

Harvard professor Edward Glaeser puts forth a few compelling models. The first one is similar to a “technical analysis” used for stocks. It maintains that a few statistical norms of the housing market will continue on into the future: Long-term reversion to the mean and short-term momentum. This first model starts with long-term mean reversion, which assumes that after big booms come big busts. Glaeser has found that for every dollar in an area’s property appreciation that same area’s prices will slide by 32 cents over the following five years. Then it evaluates shorter terms like a year or a month. In these periods, momentum prevails. For every buck in price rise per year, they will appreciate another 71 cents the next year. Glaeser attributes this in part to people’s future price assumptions being based on recent past appreciations.

When taken together, these long-term and short-term patterns create cycles. Momentum is generated at the beginning of a property rise. But this appreciation tapers off when the long-term revision takes over. Momentum still prevails, but now it is in a downward fashion.

Reassuringly, these figures seem to correlate with the most recent Case-Shiller data. (Read more about

Case-Schiller.)[br]Another forecasting approach, which also has been accurately benchmarked against the Case-Schiller hosing index, is the ARIMA time series model.

ARIMA, or AutoRegressive Integrated Moving Average, is a statistical forecasting model which fits a range of time series data. ARIMA models are often used in econometrics and statistics.

Cavemanforecaster.com has overlaid seasonal ARIMA data with Case-Schiller plots and indeed the ARIMA predicted the property drop in early 2006.

One glance at this chart and you wouldn’t believe the market was due to crash, would you?

But indeed, the forecast shows a drop (notice the red line beginning to head south).

It is not clear whether or not these data would correlate with other housing indices around the world, or whether or not this is yet a statistically reliable approach, but there’s something to be said for being able to predict a market.

Both Glaeser and Cavemanforecaster.com present compelling forecast models. But why didn’t they tell us earlier?

Hiroko Mirafiori, EconomyWatch.com