Seattle Bubble

I was going back and updating some of my lesser-used spreadsheets today when I came upon one that I thought was particularly interesting. The graph below is an updated version of something I presented originally in the post Home Buying Demand vs. Price Changes back in April.

The chart below takes the year-to-year change in the single-family median home price as reported by the NWMLS and compares it to the persons per closed (SFH) sale, using the monthly Civilian Labor Force series from Workforce Explorer. All of the data is for King County as a whole.

The green dots represent the period of increasing home price appreciation from July 2003 through January 2006, while the red dots are the data from January 2006 through November 2008, when home price appreciation has been on the decline. The unfilled white dots represent January 2000 through June 2003, when appreciation was in a state of fluctuation.

There is no time-shifting in this chart, just the rolling averages to smooth out the noisy data sets. Here’s a plot of the raw data (thin dashed lines) and the rolling averages for each set:

To be honest, I’m not entirely sure what to make of this data. I realize that the last time I posted charts similar to these some folks railed into me (rightly so) for some statistical nonsense in a few of the charts. This time I have attempted to keep things simple, and I think the results are quite interesting.

From what I can tell, there seems to be a very close relationship between the number of sales as a percentage of the population and the change in price. The strange thing to me is how different the slopes are on the red and green lines.

I would postulate that the closely clustered white dots represent the relationship between sales volume and price changes in a “normal” or “balanced” market, while the dramatically-sloped green dots are a result of the out-of-whack mentality of the bubble market, and the red dots are the result of a correcting post-bubble market.

So what do you think? I know there are quite a few readers out there that are more statistically-inclined than myself, and I’d love to have some of you download the data for yourself and give me your take on it.

In theory, there are two factors that affect the price of homes: supply and demand. We’ve looked extensively at the relationship between supply (inventory) and price in the past. Let’s take a look at the relationship between demand and price.

For the purpose of this post, we will measure demand by looking at the relationship between the number of closed sales in a month and the total population. For population, I’ll be using the “Civilian Labor Force” data from Workforce Explorer Washington, since it is reported monthly. Note that the number I’m using is not the number of people employed, but the total number of employable people. For the median price and total number of closed sales, I’ll be using the single-family home data released monthly by the NWMLS. All of the data will be for King County as a whole.

First, let’s have a look at a raw chart of all the data, which is available through early 1999:

Click to enlarge

In order to keep the graph from being an unintelligible mess, I’ve graphed the “1 Sale per X People” as a 12-month rolling average. This smooths out the large spikes that occur due to the highly seasonal nature of home sales. The YOY price change is also a rolling average, but only 6 months was necessary to smooth it out. You can see the raw data for both series in faint dashed lines.

Just by looking at this graph, you can see that there seems to be a relationship between the two—when the number of people per closed sale decreases, the price changes increase, and vice versa. Let’s take a closer look at this by graphing the two running averages on a scatter plot.

Click to enlarge

Clearly there’s some sort of relationship going on here, but with an R² of just over 0.5, it’s not very strong. Let’s take a page out of Deejayoh’s playbook and see what it looks like if we compare each month’s rolling-average sales data with price change data sometime in the future. I looked at 3, 6, 9, and 12-month delays, and found that the strongest relationship was in a 9-month delay:

Click to enlarge

With an R² of 0.81, now we’re talking. But what’s with that trail of dots (that I have highlighted in green) deviating so severely from the pattern of all the rest? Those represent the YOY price change data from the last 6 months, October through March. If we stop the series in September, the R² jumps up to over 0.9.

So clearly there was a strong relationship between demand and home price changes, at least until late last year, when things began to fall apart.

But hold on a minute. Let’s go back to that first scatter plot again. I’ve highlighted the last six data points again in green, and given them their own trend line:

Click to enlarge

Whoa. Granted, 6 months of data isn’t much to go by, but still, R² of nearly 1.0 is pretty hard to ignore. I think this is definitely a trend to keep an eye on. If we make the fairly reasonable assumptions that population will continue to grow at the average rate it has grown the last 12 months and YOY sales will continue to drop at the average rate of the last 6 months, this trend line would result in YOY median price drops approaching 20% by the end of 2008.

I am not saying that is what will happen, although it certainly could. I just find it interesting that the time-delay in the relationship between demand and prices seems to have all but vanished with the recent changes in the housing market. Who knows how long it will continue, and who knows what population and sales will really do. What I do know is that I will definitely be paying close attention to this relationship as the mess continues to unfold here in Seattle.

Sources:
Sales & Prices: NWMLS
Labor Force: Workforce Explorer Washington