Estimation

The estimation component generates estimates of fair market value (FMV) for vacant and “mostly” vacant properties in the contiguous United States.

FMV estimates are predictions (conditional means) produced by statistical models trained on samples of millions of private property transactions. Models are trained at various geographic scales using simple estimators (linear regressions, tree ensembles) and publicly available predictors.

We also estimate associated uncertainties and identify the geographic regions for which these uncertainties are assumed to be valid (areas of applicability).

• Samples
  • Geographies
  • Parcel types
  • Sale types
• Models
  • Estimators
  • Predictor sets
  • Model specifications
• Uncertainty
  • Model-level uncertainty
  • Parcel-level uncertainty