The Great Miami Condo Head Fake: How 125 sales duped forecasting models far and wide

August 30, 2010

by Scott Sambucci

6 comments

Examples like these show how vital it is to look at active market data when forecasting local real estate markets.  If your model accounts only for transaction prices, you’re costing yourself (and your investors) lots of money with a really bad and latent forecast.

Check out median sold prices for condos in Miami, FL (based on transactions from county records).  In late October 2009 through early January 2010, sure looked like the condo market was recovering down in Miami.  Notice the 15% spike in prices from just under $140k to $160k in the Fall:

Miami Condo Sold Prices, Jan 2009-Jan 2010 (90-day rolling average of weekly observations)

So what exactly happened here? First, let’s take a look at market prices by quartile, defined as 25% price bands based on active market properties. There is a noticeable gap between the most expensive 25% of condos for sale (median price = $500k) as compared to the remaining 75% (median price < $250k):

Miami Condos: Median Ask Prices by Quartile

Next, let’s overlay the active inventory counts (black line, left axis) and the transaction counts (orange line, right axis). Notice how there are 8000-9000 condos available for sale each week, while actual transactions huddle around 175 per month during the October 2009 to January 2010 period, with a jump to 275 transactions for a single week in mid-November:

Miami, FL: Active Listings vs. Sold Count, Fall 2009 (weekly observations)

Now, let’s combine our quartile analysis with transaction counts to view “sold count by quartile” per week during this period.  Notice the single-week spike in mid-November – there were 125 condos sold in the top price quartile (blue line below = condos priced in the $500k range):

Miami Condo Sold Counts by Quartile

Weekly transactions by Price Quartile

And finally, let’s see how the Miami Condo market is really playing out in 2010 (hint: it’s not):

Miami - Condo Sold Prices (90-day rolling average of weekly observations)

Miami Condos: Sold Prices through May 2010 (90-day rolling average of weekly observations. 90-day RA values are typically used in forecasting models)

All the while, what did real-time active market indicators (Median Ask Price and the Price of New Listings) reveal?  That the Miami Condo market was not, in fact, improving over the longer run, and certainly not heading into recovery mode:

Miami Condos: Real-time Market Ask Prices vs. Sold Prices

The single-week high-end transactions influenced 90-day rolling average sold prices into early January 2010, drastically altering 2010 forecasts.

What’s the lesson here?  In a city with 500,000 people and 9000 condos available for sale, 125 high-end transactions during the week of 11/15/2009 duped the transaction-based home price forecasts.  Additionally, remember that transaction price availability lags by at least a couple of months vs. real-time, active market indicators. This left traditional models using limited 2009 data, which came available around March 2010, indicating that the Miami Condo market was on the rise. Meanwhile, the rest of us using real-time active market data, including the thousands of remaining sellers still on the market, all knew that the one-week sales spike back in November was an aberration.

Let me know the next time you’re trading on a transaction price model. I’d like to be on the other end of that trade.

Previous post:

Next post: