patrick.net

FunTime says

The thinking you've communicated is the same I use to think that once a lot of people have hunkered down in their houses, prices can still go way below what most people paid because those of us left are not willing to spend 50% of our income buying a house that will return only a major hobby and living in a house and the conforming behavior of saying, "I own a house."

Right, prices are set by actual transactions happening.

JH says

And that is exactly why medians work. Because both happen, highs and lows, and in the long term the median accounts for both.

That's not quite true. Let's use an extreme example on both sides to show how medians can create distortions in the data. In Zip Code #1, 6 houses sell for $100K and 4 houses sells for $1 million. Median house price is $100K. In Zip Code #2, 4 houses sell for $100K and 6 houses sell for $1 million. Median price is $1 million.

Are those zip codes that fundamentally different?

Zip Code #3 has 99 houses sell for $100K and 1 house sell for $1 million. Median is $100K. Is that zip code exactly the same as Zip Code #1 because the median is the same?

Does this mean medians are completely useless? Absolutely not. But you have to consider it as part of the overall data. Analysis of data has to be more holistic than just looking at one number.

Mean is similar. It matters whether the mean is $400K with a standard deviation of $20K vs. a standard deviation of $100K.

This is why I like looking at raw data too. You can see if there are two peaks or if there are other anomalies in the data. You can see if people who buy cheap houses are not in the marketplace vs. people who buy expensive houses.

All I can say is that at least around here in the bay area real estate AND rents are insanely high and in my opinion, neither option makes any sense at all. This all changed dramatically. We bought 2 years ago and as of now if we were to for some reason sell, renting anything around here would be a lot more than the current mortgage. Buying would set us back an additional 30%+ for the same house.

edvard2 says

We bought 2 years ago and as of now if we were to for some reason sell, renting anything around here would be a lot more than the current mortgage. Buying would set us back an additional 30%+ for the same house.

Yeah, the area went totally insane, again, in that time. It is insanity, though. No data suggests to me that people can really afford it. They're just going all-in. Good luck, player!(not meant for you edvard2)

corntrollio says

In Zip Code #1, 6 houses sell for $100K and 4 houses sells for $1 million. Median house price is $100K. In Zip Code #2, 4 houses sell for $100K and 6 houses sell for $1 million. Median price is $1 million.

Are those zip codes that fundamentally different?

I don't see this happening in the middle to upper middle class neighborhoods I was arguing initially with tatapupu about earlier on this thread. 100k vs 1 M is an extreme example, and I get your point, but more likely we are talking about 500k v 700k or 400k v 600k, along with a better distribution of data points, making medians more realistic, and your scenario less realistic.

Therefore, in the case of low inventory in very expensive areas I completely agree that raw data is best. Median v. mean doesn't tell the story...at all. But I am talking about larger trends in metro areas like San Jose, Orange County (not coastal), etc. The fact that the incomes of the top 1% of earners has increased disproportionately has very little bearing on these markets. These markets are driven by financing and therefore affordability with respect to median income...or 60-80% quintile income. NOT top 1% income.

FunTime says

No data suggests to me that people can really afford it.

People can't afford it. No analysis needed. The incomes are not even close.

JH says

100k vs 1 M is an extreme example, and I get your point, but more likely we are talking about 500k v 700k or 400k v 600k, along with a better distribution of data points, making medians more realistic, and your scenario less realistic.

Medians still aren't any better. Price point matters a whole lot and sometimes lower priced houses are leading the charge (recent housing boom) and sometimes higher priced houses are leading it a bit more (current boom seems more lead by higher priced).

Case-Shiller publishes three tiers of sales, and they all act slightly differently. Here is an example for San Francisco:

http://www.socketsite.com/archives/2014/02/san_francisco_home_appreciation_slows_condo_values_stal.html

As you can see, Case-Shiller low tier went the highest and tanked the most in the big boom and bust, and the other two tiers acted slightly differently. To argue that this doesn't change median doesn't make sense.

In some cities of the Bay Area, you can see certain price thresholds under or over which houses are or aren't selling at various points. There have been certain points where the higher priced ones have been selling more (increasing median) and other times where lower priced ones have been selling more (lowering median). Median ignores all this and assumes that everything is comparable across all time periods.

That's why Case-Shiller's methodology is generally better than pure median.

Renters are losers

Bm05211983 says

Renters are losers

.... and your mom is a whore. What your point?

corntrollio says

That's why Case-Shiller's methodology is generally better than pure median.

I have to say, though, that the middle third tracks pretty damn well in this chart. Separating into thirds helps understand the boom/bust of the mid 2000s for sure. That is where a more refined analysis that includes raw data where practical is definitely useful. However, the previous discussion which now appears to be gone is looking long-term...30 years. From your graph I can see that all thirds tracked tightly from 87 to 01. And now they are returning back to that tight tracking now in 14. Seems like the medians are returning. Hmmmm....

Oh, and that the middle third held the average/median pretty well.

Oops I liked my own comment, when I meant to click edit. haha

evilmonkeyboy says

Bm05211983 says

Renters are losers

.... and your mom is a whore. What your point?

Nice. I was hoping Patrick had dumped this BM troll.

JH says

However, the previous discussion which now appears to be gone is looking long-term...30 years. From your graph I can see that all thirds tracked tightly from 87 to 01. And now they are returning back to that tight tracking now in 14. Seems like the medians are returning. Hmmmm....

Oh, and that the middle third held the average/median pretty well.

Over the long term, some of this averages out, sure. In the short-term, different things can happen. If you zoom in on the 1989-1990 peak and the subsequent bust, you can see that it wasn't exactly all together, but it looks like it now when you zoom out to the larger picture. The lines may converge a bit going forward, they may not. We'll see.

Part of what has changed medians dramatically now vs. 2008-2011 or so, is that a lot of the foreclosures and short sales that composed more of the sales are now gone, but they had made Case-Shiller tank like crazy then. A lot more of the high-priced stuff is selling overall vs. the last few years right now.

In addition, the high flying equity market is making ultra-high-end places trade at crazy valuations right now, which is bringing everything up as people fill in the gap. The mix of housing always matters for median in the short-term, and it makes it hard to compare things across different periods sometimes too.

Note also that Case-Shiller tiers are determined by the first purchase price in the pair, not the second.