Friday, June 04, 2010

dive bombing assessments

It is Midway Day which reminded me of things going on that we have not caught up with.   For example the county is pushing forward with the court mandated mass reassessment Judge Wettick has ordered.  The latest news from last month is that the county is contracting with a firm that will use aerial photos and computer enhancement to compile property and parcel metrics.  So it's like they will be reassessing by dive bomber or something.

Ok.  bad joke I guess.  In bad taste at least.  Especially since it brings to mind the MOVE fiasco in Philly some years ago.  One level of disfunction we have not caught up to Philly on.

That and it really is Midway Day.  Any veterans from the battle still around?  I wish I knew.  Just not the same energy around here for WWII history since Goldy retired.

On assessments though. However they want to collect data is fine. My one point is that the problems of assessing property in Allegheny County has never been about the quality of the data.  The problem that is acute in Allegheny County is with the modeling of the data.  From the Sabre Systems values, to the CLT numbers compiled the first or 2nd or 3rd times they all suffer from the being unable to take the data and then put good values on properties across the region.  It results from the topography of how we live, especially patterns of segregation and similar factors that make modeling Allegheny County a lot more difficult than almost any other place in the nation. If you just run the local numbers through models used elsewhere all that will result is a repeat of less than ideal numbers that have been generated in the past. 

But it will be years before we see any new values from the assessment en train now.   Until there are some new numbers both public and media attention will be scant.  But when those numbers do come I suspect many will be quite surprised.


Anonymous MH said...

It results from the topography of how we live, especially patterns of segregation and similar factors...

Mathematically, that would be pretty easy to account for and that data is easily available. Not that I'm in favor of opening that can of worms.

Saturday, June 05, 2010 12:01:00 PM  
Blogger C. Briem said...

You'll have to trust me it's not.

To do decent modeling of prices you need data on recent sales with valid market prices. At the same time the universe of comparable properties you want to be as narrow as possible. There is this fundamental tradeoff between defining your geography as narrowly as possible, yet capturing enough recent sales to be able to model a price. that tradeoff exists in all mass assessment modeling of course, but for us the problem becomes acute for several reasons.

One problem is that how we reside as homogeneous communities can be by very small geographies.. and geographies that are not always easily definable. But even if we can it is easy to run into a problem that the geography you want to use just lacks enough data to model. Then there is an even more pathological problem that many of our communities have valuations so low that true market prices are zero if not negative. That is a deep problem.. but it plays out in a more practical problem in that in lots of places you have a fundamental problem distinguishing non arms-length transactions from actual market sales because they are so low-valued. that is actually what really did in the first Sabre systems produced numbers.

and that is all the very short story of it all.

Sunday, June 06, 2010 9:16:00 PM  
Anonymous MH said...

I'm not doubting that and I know nothing about what is allowed in those models. However, you'd mentioned segregation and I was thinking that you could (not that the county should) predict values much better if you matched to the census data on race and income for that census block and that exactly nobody would want to see the betas from that model.

Sunday, June 06, 2010 10:36:00 PM  
Blogger C. Briem said...

There is a philosophical debate over what the difference is between econometrics and statistics embedded in that comment. But we’ll skip that.

I think those main problems I identified would still be there even if one could include aggregate race stats in some form as an explanatory variable.. I am not quite sure that would be legal, and even if it were it raises all sorts of other issues. It could create a feedback loop of sorts that would be akin to endogenous redlining.

As for income.... if aggregate level income stats aren’t getting capitalized into house value in some form then there are probably bigger issues with trying to model any of this.

Sunday, June 06, 2010 10:53:00 PM  
Anonymous MH said...

I am not quite sure that would be legal, and even if it were it raises all sorts of other issues.

Clearly, it would be a very bad idea for many reasons.

Sunday, June 06, 2010 11:06:00 PM  

Post a Comment

Links to this post:

Create a Link

<< Home