1. Context of the dataset: This dataset was collected as a part of the Willamette Water 2100 project, which evaluates how climate change, population growth, and economic growth will change the availability and the use of water in the Willamette River Basin (WRB) on a decadal to centennial timescale. The five-year project began in October 2010, and is a collaborative effort of faculty from Oregon State University (OSU), the University of Oregon (UO), and Portland State University (PSU). The team is developing a computer model using the Envision platform, a computing framework developed at OSU. The model will make it possible to explore how climate change, population growth, and economic growth will alter the availability and the use of water in the WRB. Envision provides a computing environment in which state-of-the-art hydrological, ecological, and economic models interact synergistically. These data were used to estimate the land values that enter into the Land-Use Transitions modeling component of Willamette Envision, the computer model developed by the Willamette Water 2100 Project. The land values were collected for randomly drawn samples of developed, agricultural, and forest land parcels in Benton, Lane, Marion, and Washington Counties for the years 1973, 1980, 1986, 1992, and 2000. Each row in the .csv file represents a unique parcel-year combination. Note that, due to the nature of the parcel-level land value information, all parcels are not observed in every year. 2. A list of the variables in the dataset: Each row in the .csv file represents a unique parcel-year combination. Note that, due to the nature of the parcel-level land value information, all parcels are not observed in every year. These data were collected from the Benton, Lane, Marion, and Washington County assessment offices over the course of the 2011-2012 academic year and into the beginning of the 2012-2013 academic year. For each county I drew a random sample of properties from the 2000 county real property roll, which was available in digital format. The samples were stratified according to the three broad land use categories of development (residential/multi-family), agriculture, and forest. Source acronyms are defined as: County assessor's office (CA), Geographic Information Systems (GIS), Google Maps (GM), United States Geological Survey (USGS), Department of Land Conservation and Development (DLCD), United States Census of Population (USCP), Oregon State Forestry Science Lab (OSFSL), generated by original code from Dan Bigelow (DB), and Environmental Protection Agency (EPA). The variable list below contains the variable name, description, measurement units, whether the measurement varies by time, the data source, and whether the variable was used in the Willamette Water 2100 Envision Model. The list is structured as follows: [Variable name] =[Description] ; [Measurement Units] ; [Time-Varying?] ; [Source] ; [Used in WW2100 Model?] newid =numeric index of parcels ; N/A ; N/A ; DB ; No: year =1 if year corresponding to parcel value is t, 0 otherwise; t=1973, 1980, 1986, 1992, 2000 ; Binary ; Yes ; CA ; Yes: MTL =map-taxlot number for parcel ; N/A ; N/A ; CA ; No: PI =1 if parcel is under private industrial ownership, 0 otherwise ; Binary ; No ; GIS; OSFSL ; No: PNI =1 if parcel is under private non-industrial ownership, 0 otherwise ; Binary ; No ; GIS; OSFSL ; Yes: acdif = difference between GIS and assessor's acreage ; Acres ; No ; GIS; CA ; No: benton =1 if parcel is located in county c, 0 otherwise; c=Benton, Lane, Marion, and Washington ; Binary ; No ; CA ; Yes: lane =1 if parcel is located in county c, 0 otherwise; c=Benton, Lane, Marion, and Washington ; Binary ; No ; CA ; Yes: mean_elev =mean elevation of parcel ; Meters ; No ; GIS; USGS ; Yes: mean_slope =mean slope of parcel ; Degrees ; No ; GIS; USGS ; Yes: near20kmi = Euclidean distance to nearest city center of over 20,000 people ; Miles ; No ; GIS; GM ; Yes: near20ksq = squared Euclidean distance to nearest city center of over 20,000 people ; Miles ; No ; GIS; GM ; Yes: pc4 =1 if parcel included in estmated in given time-period, 0 otherwise ; N/A ; Yes ; GIS; CA ; No: poly_ac = area of parcel ; Acres ; No ; GIS; CA ; Yes: pop_den = population density of nearest city of over 20,000 people ; Thousands of ppl./sq. mile ; Yes ; USCP ; Yes: prop_type = ag (agriculture), for (forest), or dev (developed) ; N/A ; N/A ; CA ; No: wash =1 if parcel is located in county c, 0 otherwise; c=Benton, Lane, Marion, and Washington ; Binary ; No ; CA ; Yes: riv_feet =total length of river reaches running through parcel ; Feet ; No ; GIS; EPA ; Yes: efu_zone2 =1 if parcel located in an EFU zone, 0 otherwise ; Binary ; No ; GIS; DLCD ; No: for_zone2 =1 if parcel located in a forestry zone, 0 otherwise ; Binary ; No ; GIS; DLCD ; No: ugb_ti =1 if parcel located in a UGB, 0 otherwise ; Binary ; No ; GIS; DLCD ; Yes: efu_ugb_mi2 =distance to nearest UGB edge if parcel is located in an EFU zone ; Miles ; No ; GIS; DLCD ; No: for_ugb_mi2 =distance to nearest UGB edge if parcel is located in a forestry zone ; Miles ; No ; GIS; DLCD ; Yes: hhinc = average houshold income for county in which parcel is located ; $1000/household ; Yes ; WP ; No: fr_per2 =average per-acre farm rents ; $/acre ; Yes ; WW2100 Farmland rent model ; Yes: imrfor =Inverse Mill's ratio for forest parcels ; N/A ; Yes ; DB ; No: imrag =Inverse Mill's ratio for agricultural parcels ; N/A ; Yes ; DB ; No: imr_uvd2 =Inverse Mill's ratio for developed parcels ; N/A ; Yes ; DB ; No: log_hhinc_WP =natural log of average houshold income for county in which parcel is located ; log($1000/household) ; Yes ; WP ; Yes: year80 =1 if year corresponding to parcel value is t, 0 otherwise; t= 1980, 1986, 1992, 2000 ; Binary ; Yes ; CA ; No: year86 =1 if year corresponding to parcel value is t, 0 otherwise; t= 1980, 1986, 1992, 2000 ; Binary ; Yes ; CA ; No: year92 =1 if year corresponding to parcel value is t, 0 otherwise; t= 1980, 1986, 1992, 2000 ; Binary ; Yes ; CA ; No: year00 =1 if year corresponding to parcel value is t, 0 otherwise; t= 1980, 1986, 1992, 2000 ; Binary ; Yes ; CA ; Yes: impadj =inflation-adjusted improvement values on parcel ; $ ; Yes ; CA ; Yes: lnperadj =natural log of inflation-adjusted per-acre real market value ; log($/acre) ; Yes ; CA ; Yes: ugb80 =1 if parcel is in sample and inside of a UGB in year t, 0 otherwise ; Binary ; Yes ; GIS; CA ; No: ugb86 =1 if parcel is in sample and inside of a UGB in year t, 0 otherwise ; Binary ; Yes ; GIS; CA ; No: ugb92 =1 if parcel is in sample and inside of a UGB in year t, 0 otherwise ; Binary ; Yes ; GIS; CA ; No: ugb00 =1 if parcel is in sample and inside of a UGB in year t, 0 otherwise ; Binary ; Yes ; GIS; CA ; Yes: mean_hhinc2aux =parcel mean of county-level household income ; $/household ; No ; WP ; Yes: mean_imragaux =parcel mean of agricultural inverse Mill's ratio ; N/A ; No ; DB ; No: mean_imrforaux =parcel mean of forest inverse Mill's ratio ; N/A ; No ; DB ; No: mean_impadjaux =parcel mean of improvement values ; $ ; No ; CA ; Yes: mean_popdenaux =parcel mean of population density of nearest city ; Thousands of ppl./sq. mile ; No ; USCP ; No: ehatbar10 =adjustment factor for developed parcels ; N/A ; N/A ; DB ; Yes: ehatbarag6 =adjustment factor for agricultural parcels ; N/A ; N/A ; DB ; Yes: ehatbarfor6 =adjustment factor for forest parcels ; N/A ; N/A ; DB ; Yes: 3. Contact Person: Dan Bigelow (PhD, Applied Economics, Oregon State University, 2015) daniel.bigelow@ers.usda.gov 4. Methods: These data were collected from the Benton, Lane, Marion, and Washington County assessment offices over the course of the 2011-2012 academic year and into the beginning of the 2012-2013 academic year. For each county I drew a random sample of properties from the 2000 county real property roll, which was available in digital format. The samples were stratified according to the three broad land use categories of development (residential/multi-family), agriculture, and forest. To stratify each county’s population of taxlots, I used the property class code, which indicates the highest and best use for each individual parcel according to its physical characteristics and any associated zoning restrictions. The property class code also represents the land use upon which each parcel’s RMV is based. Several different property class codes comprise each of the three land uses (developed, agriculture, forest) that I consider in the analysis. To account for this I grouped the property class codes that represent each of these uses together and then randomly sampled from each land use strata and county using Stata. The resulting initial samples of taxlots in each use vary in size across the four study area counties. For Benton County, the initial sample consists of 700 parcels, 400 of which are classified as developed or developable (i.e., vacant) for residential housing and 150 each in agriculture and forest. Since it is associated with the smallest urban area represented by the study counties, the size of the Benton County sample is smaller than that of any of the other three study counties. The initial samples for Lane, Marion, and Washington Counties are proportional to those of Benton County in terms of the size of the urban area they represent. Specifically, Washington County has the largest sample overall due to its association with the Portland Metro area, followed by Marion County (Salem), Lane County (Eugene-Springfield), and, again, Benton County (Corvallis). In each county, the initial sample size for developed parcels is roughly one-third larger than the combined samples of undeveloped parcels (agriculture and forest). The number of initial forest and agricultural parcels is equal within each of the four counties, but differs across counties.