parcel

The goal of parcel is to provide tools for matching real-world addresses to reference sets of addresses; e.g., “352 Helen Street”, “352 Helen St.” or “352 helen st”. This package is motivated by the included example data resources of auditor parcel tax data from Hamilton County, Ohio.

With this specific goal in mind, parcel includes:

Installation

{parcel} requires the usaddress python module for tagging addresses and creating address stubs, as well as the dedupe and dedupe-variable-address python modules for matching addresses, all of which can be installed from inside R with:

reticulate::py_install("usaddress", pip = TRUE)
reticulate::py_install("dedupe", pip = TRUE)
reticulate::py_install("dedupe-variable-address", pip = TRUE)

The development version of parcel can be installed with:

pak::pak("geomarker-io/parcel")

Example Usage

Use get_parcel_data() to get the corresponding parcel data for a vector of addresses:

library(parcel)
get_parcel_data(c("1069 Overlook Avenue Cincinnati OH 45238",
                  "419 Elm St. Cincinnati OH 45238",
                  "3544 Linwood Av Cincinnati OH 45226"))
#> # A tibble: 3 × 23
#>   input_address              parcel_id score parcel_address property_addr_number
#>   <chr>                      <chr>     <dbl> <chr>          <chr>               
#> 1 1069 Overlook Avenue Cinc… 1800A800… 0.717 1069 OVERLOOK… 1069                
#> 2 419 Elm St. Cincinnati OH… 54000410… 0.733 419 ELM ST     419                 
#> 3 3544 Linwood Av Cincinnat… 01900010… 0.733 3544 LINWOOD … 3544                
#> # ℹ 18 more variables: property_addr_street <chr>, property_addr_suffix <chr>,
#> #   condo_id <chr>, condo_unit <chr>, parcel_centroid_lat <dbl>,
#> #   parcel_centroid_lon <dbl>, market_total_value <dbl>, land_use <fct>,
#> #   acreage <dbl>, homestead <lgl>, rental_registration <lgl>,
#> #   RED_25_FLAG <lgl>, year_built <int>, n_total_rooms <int>, n_bedrooms <int>,
#> #   n_full_bathrooms <int>, n_half_bathrooms <int>,
#> #   online_market_total_value <dbl>

Python, miniconda, and virtualenv

reticulate::py_install() assumes a non-system version of Python is already installed and will offer to install Miniconda and create an environment specifically for R and the reticulate package.

As an alternative to miniconda, it is possible to create a virtualenv using reticulate:

library(reticulate)
install_python("3.9.12")
virtualenv_create("r-parcel", version = "3.9.12")
use_virtualenv("r-parcel")
py_install("usaddress", pip = TRUE)
py_module_available("usaddress")

To help R find this virtualenv for every session, either use_virtualenv("r-parcel") should be run in every R session or, better yet, by setting the RETICULATE_PYTHON environment variable in a user- or project-specific .Renviron file. You can check on which python installation chosen to be used by reticulate and why by using:

reticulate::py_config()
reticulate::py_list_packages()

CAGIS Parcels Data

The cagis_parcels tabular data resource (TDR) is created using the R scripts in /inst and stored within the package. It can be loaded using {codec}:

d_parcel <- codec::read_tdr_csv(fs::path_package("parcel", "cagis_parcels"))

head(d_parcel)
#> # A tibble: 6 × 15
#>   parcel_id     parcel_address       property_addr_number property_addr_street
#>   <chr>         <chr>                <chr>                <chr>               
#> 1 6210024007000 210 CARRINGTON PL    210                  CARRINGTON          
#> 2 6210024009200 220 CARRINGTON PL    220                  CARRINGTON          
#> 3 6210024007100 210 CARRINGTON PL    210                  CARRINGTON          
#> 4 6210024008200 220 CARRINGTON PL    220                  CARRINGTON          
#> 5 5000122021500 8053 WITTS MILL LN   8053                 WITTS MILL          
#> 6 5000122020900 8120 WITTS MEADOW LN 8120                 WITTS MEADOW        
#> # ℹ 11 more variables: property_addr_suffix <chr>, condo_id <chr>,
#> #   condo_unit <chr>, parcel_centroid_lat <dbl>, parcel_centroid_lon <dbl>,
#> #   market_total_value <dbl>, land_use <fct>, acreage <dbl>, homestead <lgl>,
#> #   rental_registration <lgl>, RED_25_FLAG <lgl>

# without codec:
# read.csv(fs::path_package("parcel", "cagis_parcels"))
options(knitr.kable.NA = '')
codec::glimpse_attr(d_parcel) |>
  knitr::kable()
name value
profile tabular-data-resource
name cagis_parcels
path cagis_parcels.csv
version 0.8.2
title CAGIS Parcels
homepage https://github.com/geomarker-io/hamilton_parcels
description A curated property-level data resource derived from the Hamilton County, OH Auditor data distributed through CAGIS Open Data: https://cagismaps.hamilton-co.org/cagisportal/mapdata/download 
options(knitr.kable.NA = '')
codec::glimpse_schema(d_parcel) |>
  knitr::kable()
name title description type constraints
parcel_id Parcel Identifier uniquely identifies properties; the auditor Parcel Number string
parcel_address Parcel Address derived by pasting Property Address Number, Street, and Suffix string
property_addr_number Address Number string
property_addr_street Address Street string
property_addr_suffix Address Suffix string
condo_id Condo identifier used to match two parcels to the same building of condos string
condo_unit Condo unit specifies a specific unit within a building of condos string
parcel_centroid_lat Parcel Centroid Latitude coordinates derived from centroid of parcel shape number
parcel_centroid_lon Parcel Centroid Longitude coordinates derived from centroid of parcel shape number
market_total_value Market Total Value number
land_use string apartment, 4-19 units, apartment, 20-39 units, apartment, 40+ units, mobile home / trailer park, other commercial housing, office / apartment over, single family dwelling, two family dwelling, three family dwelling, condominium unit, boataminium, condo or pud garage, landominium, manufactured home, lihtc res, other residential structure, metropolitan housing authority
acreage Acreage number
homestead Homestead boolean
rental_registration Rental Registration boolean
RED_25_FLAG boolean

Some of the parcel characteristics do not make sense in certain contexts and should not be interpreted incorrectly; for example, the value of a parcel for a multi-family or multi-unit housing structure shouldn’t be compared to the value of a parcel for a single-family household for the purposes of assesing individual-level SES. Within the process of matching to a parcel, an individual address could be merged with differing types and resolutions of data:

%%{init: { "fontFamily": "arial" } }%%

flowchart LR
classDef id fill:#fff,stroke:#000,stroke-width:1px;
classDef tool fill:#e8e8e8,stroke:#000,stroke-width:1px,stroke-dasharray: 5 2;
classDef data fill:#fff,stroke:#000,stroke-width:1px;

addr(hospitalization):::id ---> hc("likely in \nHamilton County \n (by ZIP code)"):::data
addr ---> nhc("not in Hamilton  County"):::tool

hc --> inst[institutional parcel]:::id
inst -. "institution 'type' linkage\n (e.g., JFS, CCHMC, RMH)" .-> sdoh("temporary housing,\n foster care,\n low income housing tax credit"):::data

hc --> res(residential parcel):::id

res -- CCHMC \nlinkage --> hhh("home's hospitalization history \n (i.e., pedigree)"):::data
res -- CAGIS & \nODC linkage --> hce(housing code enforcement,\n public service calls, crime):::data

res -- single family dwelling --> vat("family-level SES measures \n (e.g., value, age, condition, tenure)"):::data
res -- multi-family dwelling --> lu("auditor land use type \n (e.g., two family dwelling, \n apartment with 20-39 units)"):::data

hc --> npm(not matched \nto a parcel):::tool

Hamilton County Auditor Online Data

The hamilton_online_parcels TDR is created by linking a saved scraping of the auditor’s website to the parcel identifiers in the cagis_parcels TDR.

Similarly, the hamilton_online_parcel TDR is created using the R scripts in /inst and stored within the package. It can be loaded using {codec}:

d_online <- codec::read_tdr_csv(fs::path_package("parcel", "hamilton_online_parcels"))

head(d_online)
#> # A tibble: 6 × 7
#>   parcel_id     year_built n_total_rooms n_bedrooms n_full_bathrooms
#>   <chr>              <int>         <int>      <int>            <int>
#> 1 6210024007000       1961             8          3                2
#> 2 6210024009200       1960             7          3                3
#> 3 6210024007100       1960             7          3                2
#> 4 6210024008200       1960             7          3                2
#> 5 5000122021500       1955             6          2                1
#> 6 5000122020900       1955             6          3                2
#> # ℹ 2 more variables: n_half_bathrooms <int>, online_market_total_value <dbl>

# without codec:
# read.csv(fs::path_package("parcel", "hamilton_online_parcels"))
options(knitr.kable.NA = '')
codec::glimpse_attr(d_online) |>
  knitr::kable()
name value
profile tabular-data-resource
name hamilton_online_parcels
path hamilton_online_parcels.csv
version 0.8.2
title Hamilton Online Parcels
homepage https://github.com/geomarker-io/parcel
description A curated property-level data resource derived from scraping the Hamilton County, OH Auditor Online: https://wedge1.hcauditor.org/. Data was scraped for only residential parcels in CAGIS Parcels; see homepage for details. 
options(knitr.kable.NA = '')
codec::glimpse_schema(d_parcel) |>
  knitr::kable()
name title description type constraints
parcel_id Parcel Identifier uniquely identifies properties; the auditor Parcel Number string
parcel_address Parcel Address derived by pasting Property Address Number, Street, and Suffix string
property_addr_number Address Number string
property_addr_street Address Street string
property_addr_suffix Address Suffix string
condo_id Condo identifier used to match two parcels to the same building of condos string
condo_unit Condo unit specifies a specific unit within a building of condos string
parcel_centroid_lat Parcel Centroid Latitude coordinates derived from centroid of parcel shape number
parcel_centroid_lon Parcel Centroid Longitude coordinates derived from centroid of parcel shape number
market_total_value Market Total Value number
land_use string apartment, 4-19 units, apartment, 20-39 units, apartment, 40+ units, mobile home / trailer park, other commercial housing, office / apartment over, single family dwelling, two family dwelling, three family dwelling, condominium unit, boataminium, condo or pud garage, landominium, manufactured home, lihtc res, other residential structure, metropolitan housing authority
acreage Acreage number
homestead Homestead boolean
rental_registration Rental Registration boolean
RED_25_FLAG boolean

Inclusion/Exclusion Criteria for Parcel Data

Auditor parcel-level data were excluded if they (1) did not contain a parcel identifier, (2) did not contain a property address number/name, or (3) had a duplicated parcel identifier.

Parcels with the following land use categories are included in the data resource and others are excluded. These were selected to reflect residential usages of parcels.

library(dplyr, warn.conflicts = FALSE)

d_parcel |>
  group_by(land_use) |>
  summarize(n_parcels = n()) |>
  arrange(desc(n_parcels)) |>
  knitr::kable()
land_use n_parcels
single family dwelling 213044
condominium unit 19754
two family dwelling 11352
apartment, 4-19 units 5659
landominium 3047
three family dwelling 1863
condo or pud garage 1063
other residential structure 875
metropolitan housing authority 744
apartment, 40+ units 625
apartment, 20-39 units 457
manufactured home 204
office / apartment over 188
boataminium 141
other commercial housing 99
mobile home / trailer park 40
lihtc res 25

Non-Residential Parcels

Known non-residential addresses will be matched and returned with a special parcel identifer denoting that the matched parcel is non-residential; e.g., Cincinnati Children’s Hospital Medical Center, Jobs and Family Services, Ronald McDonald House):

c("222 E Central Parkway Cincinnati Ohio 45220",
  "3333 Burnet Ave Cincinnati Ohio 45219",
  "3333 Burnet Avenue Cincinnati Ohio 45219",
  "350 Erkenbrecher Ave Cincinnati Ohio 45219") |>
  get_parcel_data()
#> # A tibble: 4 × 23
#>   input_address              parcel_id score parcel_address property_addr_number
#>   <chr>                      <chr>     <dbl> <chr>          <chr>               
#> 1 222 E Central Parkway Cin… nonres-j… 0.620 <NA>           <NA>                
#> 2 3333 Burnet Ave Cincinnat… nonres-c… 0.733 <NA>           <NA>                
#> 3 3333 Burnet Avenue Cincin… nonres-c… 0.720 <NA>           <NA>                
#> 4 350 Erkenbrecher Ave Cinc… nonres-r… 0.733 <NA>           <NA>                
#> # ℹ 18 more variables: property_addr_street <chr>, property_addr_suffix <chr>,
#> #   condo_id <chr>, condo_unit <chr>, parcel_centroid_lat <dbl>,
#> #   parcel_centroid_lon <dbl>, market_total_value <dbl>, land_use <fct>,
#> #   acreage <dbl>, homestead <lgl>, rental_registration <lgl>,
#> #   RED_25_FLAG <lgl>, year_built <int>, n_total_rooms <int>, n_bedrooms <int>,
#> #   n_full_bathrooms <int>, n_half_bathrooms <int>,
#> #   online_market_total_value <dbl>

Estimating the number of households per parcel

Certain calculations needs to be weighted by households instead of parcel; e.g. “What fraction of families live near roadway in Avondale?”. We assume the following as a conservative estimate of the number of households per parcel for each land_use code:

land_use n households
single family dwelling 1
condominium unit 1
two family dwelling 2
three family dwelling 3
apartment, 4-19 units 4
apartment, 20-39 units 20
apartment, 40+ units 40
landominium 1
charities, hospitals, retir 1
condo or pud garage 1
metropolitan housing authority 1
office / apartment over 1
manufactured home 1
other commercial housing 1
nursing home / private hospital 1
mobile home / trailer park 1
single fam dw 0-9 acr 1
independent living (seniors) 1
lihtc res 1
condominium office building 0
other residential structure 0
boataminium 0

Identifiers for Parcels and Properties

A parcel_id refers to the Hamilton County Auditor’s “Parcel Number”, which is referred to as the “Property Number” within the CAGIS Open Data and uniquely identifies properties. In rare cases, multple addresses can share the same parcel boundaries, but have unique parcel_ids and in these cases, their resulting centroid coordinates would also be identical.

Because “second line” address components (e.g., “Unit 2B”) are not captured, a single address can refer to multiple parcels in the case of condos or otherwise shared building ownership. For example, the address “323 Fifth St” has six distinct parcel_ids, each with different home values and land uses:

parcel_id market_total_value land_use
14500010321 397500 condominium unit
14500010317 123000 condominium office building
14500010320 180000 condominium unit
14500010319 255000 condominium unit
14500010322 388230 condominium unit
14500010318 239500 condominium unit

In this case, a special parcel identifier TIED_MATCH is returned to denote that the address matched more than one parcel:

get_parcel_data("323 Fifth St W Cincinnati OH 45202")
#> # A tibble: 1 × 23
#>   input_address              parcel_id score parcel_address property_addr_number
#>   <chr>                      <chr>     <dbl> <chr>          <chr>               
#> 1 323 Fifth St W Cincinnati… TIED_MAT… 0.733 <NA>           <NA>                
#> # ℹ 18 more variables: property_addr_street <chr>, property_addr_suffix <chr>,
#> #   condo_id <chr>, condo_unit <chr>, parcel_centroid_lat <dbl>,
#> #   parcel_centroid_lon <dbl>, market_total_value <dbl>, land_use <fct>,
#> #   acreage <dbl>, homestead <lgl>, rental_registration <lgl>,
#> #   RED_25_FLAG <lgl>, year_built <int>, n_total_rooms <int>, n_bedrooms <int>,
#> #   n_full_bathrooms <int>, n_half_bathrooms <int>,
#> #   online_market_total_value <dbl>