Post-Boom House Price Dynamics (Stanley: school, income, prices)

[ACWhite]

Thanks for the additional info, PSUinNoVA, but I would not call what you described as Zillow's approach "controlling for quality differences." The "quality" of all properties is not identical at Time 1 to what it was at Time 2, because the properties and conditions themselves are not the same, due to improvements, additions, new builds, deterioration in the neighborhood, etc. The only thing that Zillow does, per your description and their own that is on the web, is include all properties rather than only sales. I "get" that you like this approach, but you should recognize that this is not the same as controlling for "quality" differences.

[PSUinNoVa]

I think I see where you might be getting confused. Although I don't know exactly how Zillow does this, I do know how some of the other valuation models work, and I would bet that Zillow's algorithm is quite similar...

Like I said before, what you really want to do is look at how the price of the exact same property changes over time. This is what people are generally referring to when they talk about the "price of housing." Say that you know the market value of all properties in Period 1. For a small subset of all properties, you also know the values in Period 2 (because the properties sold). If the properties that sold did not change substantively between Period 1 and Period 2, the differences in housing prices between these two periods is a very good measure in the change in the true cost of housing services.

Say that the appreciation rate for these properties was 10 percent. The estimated value of all the properties in the sample is then calculated as the value in Period 1 plus 10%. When some properties are sold in Period 3, the process is repeated and the values for all of the properties in the sample are updated. Note that at each step in the process, it is the change in the value of identical properties that are used to estimate the inter-period appreciation rate. This is the key difference between valuation models and just using mean sales prices to measure appreciation: the valuation models utilize the change in values of identical properties, whereas the average sales price approach uses the change of values of a heterogeneous set of properties to back out the price trend. The approaches can be shown to be mathematically equivalent when the mix of property types (e.g., large, small, luxury) sold in each is the same. When this condition does not hold, the index using the mean sales price approach is mathematically incorrect.

Back to the Zillow thing...

The process above (or something very similar, since it is proprietary) is where the Zestimate comes from. The second step in the construction of the Zillow index is just taking the median Zestimate of all properties in the sample. It is during this second step that the composition of all properties in the market could screw up the index. In fact, you can make a pretty good argument that Zillow should axe this step entirely in the index construction, instead reporting the index number from the first stage. I do want to emphasize that the Zillow approach, although flawed, is fundamentally both superior to as well as very different from "just including all properties rather than only sales." Moreover, the control for quality changes in such models comes from the paired-sales comparison that drives the value updates in each period.

From a statistical standpoint, the hard part of these exercises is figuring out how to group properties into submarkets that have similar appreciation rates. As this problem comes up a lot in a number of different contexts, there are tests that have been developed that you can test for homogeneity, and I would hope that Zillow is utilizing such tests.

Another complication that arises in the estimation of the valuation models is making sure that the property did not change much between Period 1 and Period 2 when it sold. This is really a data issue. If major improvements are reported in the data (e.g., because of permitting requirements), adjustments can be made to the model that guard against this sort of thing polluting the estimate of the price change. Many of the valuation algorithms also utilize outlier detection methods to drop properties that have a high likelihood of having undergone significant renovation.

Originally Posted by ACWhite

Thanks for the additional info, PSUinNoVA, but I would not call what you described as Zillow's approach "controlling for quality differences." The "quality" of all properties is not identical at Time 1 to what it was at Time 2, because the properties and conditions themselves are not the same, due to improvements, additions, new builds, deterioration in the neighborhood, etc. The only thing that Zillow does, per your description and their own that is on the web, is include all properties rather than only sales. I "get" that you like this approach, but you should recognize that this is not the same as controlling for "quality" differences.

[FSBox]

Originally Posted by PSUinNoVa

No, my argument has absolutely nothing to do with sales volume.

Ok, then your argument makes even less sense than I initially gave it credit for.

What you are essentially saying, is that the price of a home is made up of two components of the location (lot size, proximity to whatever, local demand), and the improvement (home size, style, quality, pool, etc).

This concept isn't abstract. Every single property-tax assessing government entity in the US recognizes this basic and simple concept. Anyone can visit Fairfax county's website to see the exact dollar amount has been assigned to the two components by an authority on the subject.

However, I don't see how this supports your assertion that people pay more for a McLean address vs Great Falls.

[kutra11]

Originally Posted by PSUinNoVa

Life has taught me that the correlation between wealth and intelligence is effectively zero.

I know this is off-topic, but can you please elaborate? I am in the midst of selecting a school for my son and would prefer saving all the money that otherwise I would be spending for his education if his being dumb guarantees (or even gives him a shot) at being wealthy.

[PSUinNoVa]

Yes, the concept of housing prices is very abstract, and there is a voluminous literature in economics and finance on this issue. The problem is that people don't bother to stop and think about how data should be interpreted. This is why you see so many stories in the press drawing asinine conclusions from data releases. I am not going to spend any more time rehashing this here, but the technical documentation on the Case-Shiller and FHFA index explains all of this pretty well: these documents are available free of charge online.

You seem to be confusing the price of the home (P*H), with the price of housing services (P). The location and improvement are both components of H. In this framework, when you get a better house (e.g., another bedroom) in the same market, H goes up, P stays the same, and P*H goes up. In the context of my McLean comment, think of it this way: say you have a house with a Falls Church address and a house with a McLean address. These homes are in the same market (P is the same) and they are identical in every other aspect (schools, structure, etc.) except for the address. If the home in McLean sells for more than the home in Falls Church, this indicates that people feel that the McLean address-- in and of itself-- is something that people value (H is higher because of the McLean address).

The standard value decomposition that you see appraisers utilize that you referred to below partitions a home's value into two components: the land and the improvements. To get the assessed value using this replacement cost approach, the appraiser estimates the cost of reconstructing the improvements. The land value component is generally estimated using comparable sales methods. Because improvements are priced at out cost, this approach implicitly lumps all changes in the cost of housing services (P) into the value of the land; that is, P is estimated solely through land transactions.

Originally Posted by FSBox

Ok, then your argument makes even less sense than I initially gave it credit for.

What you are essentially saying, is that the price of a home is made up of two components of the location (lot size, proximity to whatever, local demand), and the improvement (home size, style, quality, pool, etc).

This concept isn't abstract. Every single property-tax assessing government entity in the US recognizes this basic and simple concept. Anyone can visit Fairfax county's website to see the exact dollar amount has been assigned to the two components by an authority on the subject.

However, I don't see how this supports your assertion that people pay more for a McLean address vs Great Falls.

[PSUinNoVa]

Zero correlation between wealth and intelligence means that the incredibly dumb are as likely to be as wealthy as the incredibly smart; it does not mean that being dumb guarantees that he will be wealthy. I certainly do not believe that the correlation between intelligence and wealth is strongly negative, which appears to be how you interpreted the comment.

Originally Posted by kutra11

I know this is off-topic, but can you please elaborate? I am in the midst of selecting a school for my son and would prefer saving all the money that otherwise I would be spending for his education if his being dumb guarantees (or even gives him a shot) at being wealthy.

[kutra11]

Originally Posted by PSUinNoVa

Zero correlation between wealth and intelligence means that the incredibly dumb are as likely to be as wealthy as the incredibly smart; it does not mean that being dumb guarantees him a shot at being wealthy. I certainly do not believe that the correlation between intelligence and wealth is strongly negative, which appears to be how you interpreted the comment.

Right, so who are these incredibly dumb people who are also wealthy? Can you name a few that are in the public domain?

[PSUinNoVa]

Sure.

https://www.google.com/search?btnG=1...eople+magazine

Originally Posted by kutra11

Right, so who are these incredibly dumb people who are also wealthy? Can you name a few that are in the public domain?

[JD984]

Originally Posted by PSUinNoVa

If the home in McLean sells for more than the home in Falls Church, this indicates that people feel that the McLean address-- in and of itself-- is something that people value (H is higher because of the McLean address).

If people commonly make decisions in real life based on imperfect information, then doesn't the housing market function correctly by ascribing "H" value to that factor?

[PSUinNoVa]

Sure. I am not saying that the market isn't working; I am just saying that this would indicate that home buyers have a stated preference for the McLean address vis-a-vis the Falls Church address. There are a number of different ways that this premium can be interpreted. For instance, there might be something about the McLean neighborhoods that makes them more neighborly than neighborhoods in Falls Church. Alternatively, the premium could be a byproduct of people trying to simplify the home-buying process by classifying addresses as good, not-so-good, and bad based on information they glean from friends instead of trying to collect a lot of data on neighborhood characteristics. I have heard a lot of people say good things about McLean, so maybe more buyers put the "good" tag on McLean.

I have no idea what is driving the premium, I just noticed that it appeared to exist.

Originally Posted by JEB77

If people commonly make decisions in real life based on imperfect information, then doesn't the housing market function correctly by ascribing "H" value to that factor?