After a bit of refinement it's been possible to improve the price-change-to-MOI fit by looking at price changes over different periods and leading or lagging the datasets. Using Teranet, not MLS-HPI data, I have found the best fit to be as follows:
I would take away the following things from this graph, (and I will have much more over the coming months as monthly data become available):
- Teranet data lags months of inventory. This is for a few reasons, first that Teranet uses data based on closed transactions where MLS uses data timed on subject removals, which can be several months in advance of the closing date. Second Teranet uses some averaging to produce a more statistically significant dataset but results in a lag in reporting the data.
- Price changes have a seasonal component. Prices are usually firmer in the spring and weaker in the fall and this is part of the reason why 6 month price change correlations protrude from the data field.
- Most intriguing, and astute readers will have already figured this out, is that there is a strong predictor of the expected Teranet HPI several months in advance of it being reported, based solely on 3 month moving average MOI data.
- We have a strong indication of MOI about halfway through a month. That is, we need not wait for the REBGV reported month-end data -- with daily numbers being reported by Paul Boenisch and Larry Yatkowski we get a sneak-peak at the likely MOI value. Also check the comment section of vancouvercondo.info for some more analytical predictions of month-end inventory and sales.
- This graph validates that an MOI above 7 means prices are going to drop. Current MOI is approaching 10.
I'll have some more fun graphs in the months ahead.