Tax Rate Impact on Model Performance
We want to know whether including local tax rates improves the accuracy of our model. Please test this by creating a standalone script in R that compares models with various geographic indicators to see whether tax rates add additional explanatory power over other geographic controls.
- You will need to create a variable that captures the effective tax rates.CCAODATA.TAXBILLAMOUNTS contains each property's tax bill amount. By dividing this by the property's sale price, you get effective tax rates.
- Visualize these rates to be sure you have calculated them property.
- Construct your regression data. You will need a sample of sales in the South Tri. You will need some individual property characteristics - the usual suspects. You will need time variables. Both of these sets of variables need not be optimized for this prima facia case.
- Finally, you will need a battery of geographic identifiers: neighborhood, census tract, tax code, location factor (maybe), township.
- You just want to test a couple of model specifications that differ in their treatment of geography. We want to see whether effective tax rate adds explanatory power over an above the other geographic varaibles.
Please prepare a memo summarizing the results of your analysis. See this file for an example of a standalone script.
Fritz is the recipient of this report. @dfsnow will supervise this work.
Edited by Robert Ross