I currently have a micro (5 years) panel data of house price transaction. There are a total of 800 different high rise properties (name of apartment/condo) and 30000 transactions across the 5 years. However, the data is really unbalanced as residential properties have different numbers of transaction in a year. Also at times, a particular year does not have transaction of a particular property.
I am unsure as how I should proceed with the analysis. Should I code the property names before running the analysis or should I aggregate the values and independent variables by district and recode the districts instead?
The chapter for Unbalanced Panel of Badi Baltagi Book is good, also you will need to use indicator variables to avoid loosing information. Moreover, you need first to be sure that missing data is not due to self selection, i.e. the selection variable is not correlated to the idiosyncratic term of your model,
Ye Wu & Rick Radewagen of IE Business School recommended seven ways to handle unbalanced data. Please follow the link below
https://www.kdnuggets.com/2017/06/7-techniques-handle-imbalanced-data.html
REAL ESTATE VALUATION looks at the type, kind, location or comparable properties then class them into clusters. Analyze each cluster and compare one cluster to others of similar in character and location.
MODELING of data with five years with karge transaction like that may be advisable to divide the data into learning and testing data set. After ghe 30,000 transaction are classed into cluster on the basis of type, and location, you might randomly sample to creat a in-sample study in order to find an appropriate model then use the remaining data sets as testing sets to see the reliability and validity of your predictive model.
TIME SERIES is one way of classify this type of data. However, when applied to real estate or anything that is more practicable and less theoretical, times series may not be the best tool. It is advisible to see what type of data distribution you have and go from them for modeling approach. Let the data characteristic take you to modeling, do not force a model onto the data.
see for instance
https://www.econstor.eu/bitstream/10419/63162/1/323423752.pdf
https://homepage.univie.ac.at/robert.kunst/pan2010_pres_mayer.pdf
The chapter for Unbalanced Panel of Badi Baltagi Book is good, also you will need to use indicator variables to avoid loosing information. Moreover, you need first to be sure that missing data is not due to self selection, i.e. the selection variable is not correlated to the idiosyncratic term of your model,
I would recommend a simple solution, i.e. to remove the observations (houses I would assume in your case), for which you do not have a full set of data. If, as you said, some properties do not have transactions for some years (which is very logical) then these observations will not bring any information into your model for that particular year. Keep in mind that this solution works in case when the share of observations, for which there are lacks in data is relatively small; i.e., you will not loose a lot of information.
As for the second part of your question, what you should do next depends on the aim of your study. For example, if you want to have a general model that shows the determinants of house prices, I would run five cross-section models and then see if a) the same determinants are statistically significant and if b) the signs and the values of their coefficients have changed. In this case the district can be introduced as a set of dummy variables.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running