RCA Licensing, With A Broader Decision Support System

[RCA]

My experience over the past several years, is that the R coding required for a robust RCA system for appraisal, it not trivial.

1. R is a dynamic system that is always changing. Code you write today may not work tomorrow due to third party changes. Continual maintenance and debugging is required. Very few, if any, appraisers will have the time to deal with this.

• Advaned R code is required for a robust system with unit tests. R6 classes need to be used, with class definitions spread over several R scripts. Many advanced aspects of R need to be employed in the end, including integrration with C++ and other languages, as well as UI and Web interfaces, e.g. based on "Shiny."
  
  
• Nearly all appraisers and other users will require at least minor modifications to mold the code base to their needs. - And they will need help to manage and maintain these modifications

2. There are many problems with data collection, cleaning, maintenance and transformations that have to be tailored to individual needs.

3. The dream of employing so-called Ph.D.s to build statistical models for different market areas is unrealisitic. There will for the forseeable future only be a small cadre of appraisers who are capable of using R/MARS tools, even if such tools are supported by a third party.

• There are over 1,000 cities in the US with a population over 30,000. (multiply this by the number of property types)
• There are 3,143 counties and parishes in the US. (multiply this by the number of property types)
• Models would have to be updated at least once a year, and probably at least quarterly for most areas.
• Models have to be maintained and made available by effective date range.
• Models have to be created for about 20 years back into the past - at varying intervals depending on the market area.
• --- And it will be a very long time, if ever, before a sufficient number of competent analysts would be available to undertake such work.
• --- With the constraining influence of a very small customer base to support paying the data mining/statistical analysts the compensation they would expect.

I imagine myself, charging a license fee of about $800/year to provide limited support to an appraiser using a provided R/MARS system. They will get all of the code to use as they see fit. Plus they will get limited free support during the year. They will have to know R well enough to use it and make minor modifications to it, and of course they will have to know enough about MARS to be able to competently use it for appraisal.

In the long term, I see an RCA system the evolves to support:

• Appraisal
  • SFR, Condos (SCA- Sales Comparison Approach)
  • Duplexes, Triplexes, Quadplexes. (SCA)
  • Multi-family (SCA)
  • Other commercial RE (SCA)
  • Equipment (In particular, high-end expensive equipment such as robots and automated machinery)
• Decision Support/Consulting
  • Planning & Construction

While producing Price Models for market areas and property types seems like a good idea, the unfortunate reality is that you need domain experts who have a detailed understanding of the market, as well as competence in using R and MARS. These people will be few and far between.

Also, such a system has to be designed to exist, that is be maintained and evolve, for decades - in any case, beyond the life of any one person. So, there must be a plan in place to ensure its continued existence.

Of course appraisers, using a provided R/MARS system, could make and market their own price models, - for their market areas. I don't know if and when that would make sense. As for their own foreseeable future - they would probably be the only ones capable of using the MARS models they create for their own market areas. But hard to say.