https://www.urbantoolsconsult.org/

This link could be of interest to you, with respect to land value taxation.

If your purpose is mass appraisal of properties, then submarket is a much broader concept than what is usually done in real estate appraisal. There are much more variations within a submarket than allowed in real estate appraisal. A more practical approach might be to do appraisal along the streets and use some simple methods to do property price adjustments. Governmental appraisal of properties in an area can take a long time, for example several years. So, there is no need to rush. Of course, there is always the tradeoff between quality and time.

1. Run a hedonic regression for the entire market. 2. Try one with dummy variables that allow the implicit prices (i.e., the regression coefficients) to change with the submarkets that you think exist. 3. Are the variables with the dummies significantly different from zero? 4. If so, you have found submarkets that matter. 5. Run your hedonics for each of these submarkets separately.

Real estate markets are basically local in nature. Local demand and supply factors exert significant influence on price. It would be very effective if the appraisal is carried out using hedonic price modelling based on submarkets and according to use types. This will isolate the key determinants of price movements across all the submarkets under study.

Stephen I. Ternyik Thanks for your answer. The weblink is very useful.

Feng Frederic Deng Thanks for your opinions. Yes, I do think the block/street level is suitable for modeling. But for the submarket, the implied meaning could be the properties with similar values or variables contains the relevant submarket. The way we classify the real estate market into a submarket is the way we do the mass appraisal.

Peter F. Colwell Thanks for your kind reply. The specific research steps are very helpful and inspirational to me. I may have corresponding comments or suggestions for improvement after I do further research. Thanks again.

Namnso Bassey Udoekanem I agree with your opinion. But I'm still thinking about it. Whether the creation of submarkets will lose the integrity of the mass appraisal. After all, for a city's real estate tax evaluation, the most important thing is to be fair and transparent on the basis of accuracy and efficiency.

Daikun Wang Isn't that an example of the tail wagging the dog? You should want the best estimate possible.

Peter F. Colwell Certainly, the best estimate possible is always the eventual aim. And it's better to avoid "case of the tail wagging the dog".

There are many prospective ML/cluster-type analyses which could be employed. If one has sufficient time-series data then a Phillips & Sul convergence club approach may provide insights. As an example, refer to the NBP working paper by Matysiak & Olszewski https://www.researchgate.net/publication/334221703_A_Panel_Analysis_of_Polish_Regional_Cities_Residential_Price_Convergence_in_the_Primary_Market

What constitutes a submarket? Is it an area in which there are physically similar properties? Alternatively, is it an area in which the implicit prices are similar. I think that it is the latter. However, some others think that it is the former. You decide.

Agreed. The PS procedure can use convergence in prices, for example, as a criterion for 'similar'.

Maurizio d'Amato Yes, Cluster Analysis (CA) and Factor Analysis (FA) are helpful.

George Matysiak Noted with many thanks. I'll read your working paper.

Regarding location, my humble opinion is that GWE (geographically weighted regression) is the most effective way.

Thomas Dimopoulos Yes. I think there are mainly two types of housing evaluation, the classification (submarket) and prediction (regression). GWR model or GTWR model is an effective local regression method for prediction.