Case-Shiller vs Wages Inflation

[anforowicz]

[rose-colored-coolaid]

Nice work. It's almost like there is a disconnect that happens around 1998 between prices and wages. Funny, huh?

[deejayoh]

You compared Seattle proper to Case-Shiller - which covers all of Pierce, King, and Snohomish counties. A better comparison is to use the MSA income data vs. Case Shiller if that is the metric for home prices you want to use. I'd get BLS data for the MSA and see how it compares. Also, you might want to use a correlation vs a straight comparison of raw growth rates. It is possible that with more disposable income, home prices could rise faster than wages and a correlation would show this.

I did an earlier analysis where I correlated median home price for King County with income growth back to 1985 - and the results looked a bit different than what you show. The divergence doesn't really start until ~2003/04. Up until then incomes and home price were ~99% correlated.

[jack_carter]

I was trying to look up if and how the Case-Shiller index accounts for teardowns/sub-divisions and re-models. Given that in both seattle and in the close in neighborhoods of the eastside, most new houses are either built on tear-downs or in subdivisions, if this were left out, the index would be skewed since the new house would be 3x the old house in a year or two.

Also, since a large percentage of houses have been remodelled, without knowing the expense of the remodel, the house price increase would not be accurate.

Does anyone know how the index accounts for these?

[deejayoh]

[anforowicz]

I tried plugging national data into the spreadsheet and the chart looked more-or-less the same (houses following wages until 1997).

I tried to get the data for wages in Seattle-Tacoma-Bellevue, WA MSA. I only found data from 2001-2007 on BLS. I wish I knew how to send a direct link to BLS pages I tried. I think their AJAX-y interface doesn't allow that though, so I'll just say that I went to http://www.bls.gov/data/ and tried to get "State and County Wages" (starts in 2001) and "Earnings - National" (can't break it down into state/county/metropolitan area) and "Weekly & Hourly Earnings" (same problem - only national level) and "National Compensation Survey" (data starts in 1997).

I also went to your blog post at http://seattlebubble.com/blog/2008/01/0 ... es-to-fall but I didn't see a direct reference to your data sources (you had a link to http://www.economicforecaster.com/, but I didn't see a way to get data from that webpage).

I also wanted to compare wages to housing before 1990. CSI data is unavailable. Is there a way to get NWMLS data for (say) 1980-2007?

[deejayoh]

[anforowicz]

Thanks for reminding me about OFHEO!

Below is the chart comparing:
[list=]
[*]wages for Seattle urban workers from http://www.seattle.gov/financedepartmen ... Annual.xls (arbitrarily reindexed wages so they are closer to 2 at the beginning of the chart)
[*]OFHEO for Seattle-Bellevue-Everett, WA (MSAD) from http://www.ofheo.gov/media/hpi/1q08hpi_cbsa.csv (OFHEO: reindexed to be equal to wages index on 1980-01-01 and OFHEO2: reindex to be equal to wages index on 1990-01-01)
[*]CSI for Seattle area from http://www2.standardandpoors.com/portal ... 0,0,0.html (reindexed to be equal to wages index on 1990-01-01)
[/list]

Notice how wages and home prices follow one another in 1975-1989 and 1991-1997. I guess period 1989-1991 could be viewed as an argument against my theory that housing prices should follow wages in long-term. Perhaps something happened in 1989 that made housing more desirable or that made it easier to buy a house.

I'll try to play with the wages data from Conway-Pedersen when I get back from vacation.

home vs wages2.PNG (14.49 KiB) Viewed 5286 times

[anforowicz]

Before getting more data on wages, I wanted to look at the correlation between wages and OFHEO from the previous post.

R-squared for 1975-1996 is 0.9322. We can see on the chart that OFHEO numbers start to grow faster after year 1997 (pink dots vs trendline/blue dots).

The formula of the trendline (y=1.5867x-2.1256) can be used to calculate the expected value of the OFHEO in 2008 (assumming the trendline from 1975-1996 should extend all the way to 2008). The expected value is 57% of the actual value, so based on that data we could expect housing to fall 43%. Of course that data will change if we base the trendline on a different timeline: assumming that things were "normal" up until 2000 means that prices will fall 40%; "normal" up until 2008) means that prices will fall 22%.

correlation-wages-ofheo.PNG (18.14 KiB) Viewed 5046 times

[deejayoh]

That seems pretty consistent with what I found - but I don't expect prices to just drop in a day - so if you average the drop over 2-4 years (which is normal timespan for a downturn) and at the same time wages continue to rise during that period at a rate of 4 or 5% (which is also normal) - then the expectation for nominal price drops is more in the range of 15 -20% IIRC.

That's kind of where I come out. I think we are 1/3 to 1/2 the way down. And it's going to be slow.

Caveat: I think I tossed the pre-1980 data . It really threw off the fit and I didn't trust the numbers because of high inflation in that time.'
I've done this analysis with both wages and disposable income and come out with similar conclusions. Garth has suggested using HHI, which is a good idea - but I don't know where to find a good time series for that, and I suspect it is 99% correlated with wages anyway so there won't be a ton of new insight.

[perfectfire]

Where do you get data for a normal length of a downturn? CalculatedRisk always suggests that real estate crashes take 5-7 years to reach bottom.

[anforowicz]

Initially I picked 1996 as the last "normal" year by just looking at my first graph. If one wants the highest correlation coefficient between wages and house prices then 2002 is the best choice as the last "normal' year (at least based on the data I have in my spreadsheet):

1996 - 0.96551
1999 - 0.97383
2000 - 0.97435
2001 - 0.97484
2002 - 0.97490 <---
2003 - 0.97483
2004 - 0.97235
2005 - 0.96485
2006 - 0.95174
2007 - 0.94322
2008 - 0.93957

So below is the trendline based on wages vs prices in 1975-2002 and another graph that shows actual house prices vs prices expected based on the linear regression (aka trendline) on the first graph.

An interesting thing to note is that changing "the last normal year" from 1996 to 2002 pushes the left side of the trendline down which then shows that houses in 1975 where more overpriced than today. Interesting.

chart2.PNG (6.28 KiB) Viewed 4941 times

chart3.PNG (1.59 KiB) Viewed 4940 times