Since “inheriting” this blog in the middle of May, I’ve used the blog as an excuse to analyze data, crunch numbers and see what the results are. It’s geeky, but fun (at least for me). I’ve gained some new insights – such as when interest rates drop, prices in Victoria may – but do not always go up (outlined in my Housing prices follow mortgage rates? post).
A few years ago, Leo developed a model for predicting house sales for each month. It used the current month in the previous year, the previous month and the current month as variables to predict the sales at the end of the current month. This model works surprisingly well most of the time … except for this spring where we have a strong seller’s market and sales are not following “normal” trends. Although there are predictable seasonal variations (spring peaks, summer doldrums, winter blues) – other factors such as: interest rates, CHMC policies, and public perception of economic “health” – make it so that some months do not follow the same pattern as the previous year(s).
Therefore, I’ve been determined to try to develop a new model for predicting house sales. I have developed three variants of a new model – all based on the average daily sales for a given week. Long weekends (with statutory holidays) often skew the numbers as there can be a surge/drop in interest and delays in finalizing sales. The best model I’ve found so far is to average the previous three weeks of daily sales and to use this to predict the month-end sales. Using this model, the predictions vary each week of the month, with the prediction during the first week being surprisingly accurate:
|Month||1st Week||2nd Week||3rd Week||4th Week||Actual|
Here’s the details for July 2015 sales so far …
So why put this effort into predictions? For myself, it’s mainly an intellectual exercise. Hopefully, for you – the readers – it will help be a guide as to whether it is potentially a good time to buy or sell.