Predictors

This is the full list of parcel-level variables used as predictors and sampling filters for our models of fair market value (FMV).

These variables are derived from geospatial parcel boundaries and publicly available geospatal datasets.

In addition to the variables listed here, we obtain some predictors from parcel boundaries (geolocation and size: lat_id, long_id, ha) or transactions (date of sale).

Terrain

elev

Average elevation of parcel (meters).

Source:

USGS National Elevation Dataset (NED) 1/3 Arc-Second

Geoprocessing:

Mean of pixel values within parcel (zonal statistics). Elevation raster exported at 0.00449 degrees resolution from Google Earth Engine (EPSG:4326).

slope

Average slope of parcel (degrees).

Source:

see elev

Geoprocessing:

Mean of pixel values within parcel (zonal statistics). Slope computed at 30m resolution in Google Earth Engine (EPSG:5070).

Climate

clim_ppt_summer

Average precipitation in summer (Jun-Aug) (mm)

Source:

PRISM monthly climate normals

Access:

https://prism.oregonstate.edu/

Accessed:

Aug 10, 2023

Geoprocessing:

Zonal statistics (mean)

clim_ppt_winter

Average precipitation in winter (Dec-Feb) (mm)

Source:

see clim_ppt_summer

clim_tmean_summer

Average temperature in summer (Jun-Aug) (C)

Source:

see clim_ppt_summer

clim_tmean_winter

Average temperature in winter (Dec-Feb) (C)

Source:

see clim_ppt_summer

Hydrology

cst_50

Percentage (0-100) of coastal waters within a 50m radius. Used as proxy for beachfront properties and boating access.

Source:

ESRI North America Water Polygons

Accessed:

Jun 18, 2019

cst_2500

Percentage (0-100) of coastal waters within a 2500m radius. Used as proxy for ocean proximity for near-ocean properties. Positively associated with distance to coast as well as with the added value of properties surrounded by coastal waters on several sides, such as islands, peninsulas, etc.

Source:

see cst_50

lake_dist

Distance (m) to nearest large lake (> 4ha).

Source:

National Hydrographic Database (NHDPlus High Resolution)

Source:

National Hydrographic Database (NHDPlus High Resolution)

Access:

https://www.usgs.gov/national-hydrography/nhdplus-high-resolution

Accessed:

Jun 18, 2019

lake_frontage

Approximate total lake frontage of the parcel (in meters).

Source:

see lake_dist

Geoprocessing:

Area of intersection of parcel polygon with 50-meter-buffers around NHD lake waterbodies, divided by the buffer width (50m).

river_frontage

Approximate total river frontage of the parcel (in meters). Only waterbody polygons from the NHD are included (no lines).

Source:

see lake_dist

Geoprocessing:

see lake_frontage, but using river waterbodies.

water_exposure
Computation:

(river_frontage+lake_frontage ) / ha

p_wet

Percentage (0-100) of parcel area covered by wetland polygons.

Source:

National Wetlands Inventory (NWI), U.S. Fish & Wildlife Service

Access:

https://www.fws.gov/program/national-wetlands-inventory/wetlands-data

Accessed:

Jun 18, 2019

fld_fr_fath_p100

Flood risk: average meters of inundation depth within the 1% annual exceedance probability floodplain (pluvial floods).

Source:

Fathom-US Flood Hazard data (Wing et al 2018)

Access:

https://www.fathom.global/product/flood-hazard-data-maps/fathom-us/ (licensed)

Accessed:

Mar 26, 2020

Geoprocessing:

Zonal statistics (mean)

fld_fr_fath_f100

Flood risk: average meters of inundation depth within the 1% annual exceedance probability floodplain (fluvial floods).

Source:

see fld_fr_fath_p100

Soils

f_soil_<soil_class>

Fraction (0-1) of parcel area covered by soil_class.

Eleven soil class categories are distinguished (e.g. “prime” farmland, “state priority” soil, etc.). See Gold et al (2023) for a description.

Source:

SSURGO

Access:

https://websoilsurvey.nrcs.usda.gov/app/WebSoilSurvey.aspx

Accessed:

Aug 11, 2023

Geoprocessing:

Polygon intersections

Land cover and use

p_barren

Percentage (0-100) of pixels in parcel that were “barren” in 2011.

Source:

National Land Cover Database, Year-2011 Land Cover (Edition 2014-10-10)

Access:

https://www.mrlc.gov/data

Accessed:

June 18, 2019

p_crops

Percentage (0-100) of pixels in parcel that were “cropland” in 2011.

Source:

see p_barren

p_forest

Percentage (0-100) of pixels in parcel that were “forest” (deciduous, evergreen, or mixed) in 2011.

Source:

see p_barren

p_grassland

Percentage (0-100) of pixels in parcel that were “grassland” in 2011.

Source:

see p_barren

p_pasture

Percentage (0-100) of pixels in parcel that were “pasture” in 2011.

Source:

see p_barren

p_shrub

Percentage (0-100) of pixels in parcel that were “shrubland” in 2011.

Source:

see p_barren

irr_2000_2020

Percentage (0-100) of pixels in parcel that were “irrigated” between 2000 and 2020 (averaged across all years)

Source:

IrrMapper Irrigated Lands, Version 1.2

Access:

https://developers.google.com/earth-engine/datasets/catalog/UMT_Climate_IrrMapper_RF_v1_2

Accessed:

April 11, 2022

Buildings

All of the following indicators are derived from Microsoft’s open-source USBuildingFootprints dataset, which contains polygons of 125.2 million buildings inferred from high-resolution satellite imagery with neural networks.

Access:

https://github.com/microsoft/USBuildingFootprints

Accessed:

Aug 21, 2023

Microsoft’s building footprints are our preferred open-source metric for the presence of buildings in CONUS, as they are more consistently available across CONUS than other indicators (e.g., tax assessor data). However, building footprints introduce new sources of error. For instance, footprints under trees are often missed.

Alternative measures of building presence are available in tax assessor and parcel boundary datasets. However, their availability and quality varies across states and counties. For a comparison of ZTRAX-based and remote-sensing based building variables see Nolte et al. (2023) Land Economics (Appendix Figures A14-16)

n_bld_fp

Count of building footprints on the parcel.

Geoprocessing:

Polygon intersections.

m2_bld_fp

Area of building footprints on the parcel (square meters)

Geoprocessing:

Polygon intersections.

p_bld_fp

Percentage (0-100) of the area of the parcel that is covered by footprints.

Geoprocessing:

Polygon intersections.

p_bld_fp_*

Percentage of area within the given radius (*, integer, in meters) that is covered by building footprints. An indicator of nearby building density.

Geoprocessing:

rasterization of building footprints, pixel-based computation of average building footprint presence within circular neighborhood (2D convolution with moving-window kernel), averaged across all pixels within each parcel (zonal statistics).

p_bld_fp_500

% building footprints within 500m

See p_bld_fp_*

p_bld_fp_5000

% building footprints within 5000m

See p_bld_fp_*

Demographics

hh_inc_med_bg_2012_2016

Median household income at the census block-group level (2012-2016)

Source:

American Community Survey, via the National Historical Geographic Information System (NHGIS)

Access:

https://www.nhgis.org/

Geoprocessing:

spatial joins of parcel centroids with reference units.

p_asian_bg_2012_2016

% population in block group identifying as “Asian” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

p_black_bg_2021_2016

% population in block group identifying as “Black or African-American” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

p_hispanic_bg_2021_2016

% population in block group identifying as “Hispanic” on American Community Survey.

(Note: overlaps with ‘race’ categories, such as white, black, asian, etc.)

Source:

see hh_inc_med_bg_2012_2016

p_mixed_bg_2021_2016

% population in block group identifying as “Mixed” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

p_native_bg_2021_2016

% population in block group identifying as “American Indian or Alaska Native” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

p_pacific_bg_2021_2016

% population in block group identifying as “Native Hawaiian or Other Pacific Islander” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

p_white_bg_2021_2016

% population in block group identifying as “White” on American Community Survey.

Source:

see hh_inc_med_bg_2012_2016

bld_pop_exp_c4

Population gravity (experimental).

A spatial measure of residential population, attributed to building footprints.

Geoprocessing:

Zonal statistics

Find out more:

Infrastructure

rd_dst_pvd+

Distance to nearest paved road, including highways (meters).

Source:

TIGER/Line shapefiles from the U.S. Census Bureau for the year 2019

Access:

https://www.census.gov/geographies/mapping-files/time-series/geo/tiger-line-file.html

Accessed:

Sept 10, 2020

Only computed up to 3km.

travel

Travel time to major cities (minutes), ca. 2000

Source:

European Commission & World Bank (Nelson 2007)

Access:

https://forobs.jrc.ec.europa.eu/products/gam/

This dataset was computed with different specifications than travel_weiss. The two are not intercomparable. Differences do not necessarily reflect change over time.

travel_weiss

Travel time to major cities (minutes), ca. 2015

Source:

Weiss et al. 2017 Nature

Access:

https://www.nature.com/articles/nature25181

Land protection

p_prot_2010_5000

See p_prot_*_*

p_prot_*_*

Percentage of area within a given <radius> (in meters) that is protected by fee or conservation easement in a given <year>.

Sources:
  • Protected Area Database of the United States (PAD-US 2.0)

  • National Conservation Easement Database (NCED)

  • New England Protected Open Space (NEPOS) database

  • Colorado Ownership, Management, and Protection (COMaP) database.

Geoprocessing:

Rasterization of protection polygons, pixel-based computation of average protection within circular neighborhood (2D convolution with moving-window kernel), averaged across all pixels within each parcel (zonal statistics).

Note

Data for Colorado is licensed from COMaP and cannot be shared.

p_e

Percentage (0-100) of parcel overlapping with a conservation easement.

Sources:

see p_prot_*_*

ct_p

Percentage (0-100) of parcel overlapping with a public land acquisition.

Source:

Conservation Almanac (Trust for Public Land)

Access:

https://conservationalmanac.org/

Accessed:

Sep 15, 2019

Spatial units

Spatial reference units, ordered from those with few units (U.S. states) to those with many (census block groups).

division

U.S. census division (groups of state)

state

U.S. state, identified by its two-letter Alpha code (e.g. CA for California)

Source:

Census Bureau, via the National Historical Geographic Information System (NHGIS)

Access:

https://www.nhgis.org/

Geoprocessing:

Spatial intersection with parcel centroids

region_id

Region identifier.

Core-based regions are an experimental geographic identifier developed at the PLACES lab.

Regions divide the contiguous U.S. into less than 1000 spatial units that are identified by their high-value “core” (city centers, resorts).

We prefer modeling at the level of regions rather than counties or states, as the latter vary substantially in size and number across the U.S. geography.

Geoprocessing:

Spatial intersection with parcel centroids

Learn more:

fips

U.S. county, identified by its five-digit county FIPS code (e.g. 06037 for Los Angeles county, California)

Source:

NHGIS (see state)

zip_id

ZIP code, 2016

Source:

NHGIS (see state)

tract_id

Census tract identifier, 2016

Source:

NHGIS (see state)

bg_id

Census block group identifier, 2016

Unique within county.

Source:

NHGIS (see state)