2020 Redistricting Data for Kentucky at the District and Block level (Prison Adjusted)

These data sets include adjustments to the population to account for prisoner reallocation based on zip code of last known address prior to incarceration. The total population adjusted for returning prisoners to zip code is based on the block’s proportion of the zip code’s population. Race adjustments are based on the racial makeup of the prison population. We assign each incarcerated resident to have the same racial demographics as the state’s state-incarcerated population as a whole. For example, if a state’s state-incarcerated population is 50% white and 50% black and one block has two returning residents, we include them as 1 white and 1 black returning resident (this will bias the data toward finding a smaller racial disparity in the impact of prison gerrymandering). Our estimates may be less precise because we only adjust for state facilities and not other levels of incarceration (e.g., local jails, federal facilities). Those with out of state (n = 176) and discontinued zip codes (n = 100) were excluded from the analysis. 

Additionally, our estimates are adjusted based on ZIP code of last residence, as opposed to block of last residence, which could affect the accuracy and/or precision of the estimates. District data includes separate tabs for house and senate districts.

##Sources
Population data were retrieved from the Census Bureau, table P1 (Race) at https://data.census.gov/cedsci/all?q=P1

Prison addresses were retrieved from the Kentucky Department of Corrections at https://corrections.ky.gov/Facilities/Pages/default.aspx 

Race by prison and zip code of last known address prior to incarceration were obtained through a data request from the Kentucky DOC. 

2020 district data was obtained from Dave’s Redistricting site at https://davesredistricting.org/maps#state::KY    Maps acquired on June 13, 2022. If district boundaries have changed and you want an updated analysis, we encourage you to either use the provided raw data and update the district data or contact prisongerrymandering@gmail.com and we will update the data as soon as possible.
#Processing
The population data was recoded into three race groups to align with the racial categories provided by the state:
White is white alone (p1_003n)
Black is Black or African American alone and all multiple race categories that include Black or African American (p1_004n p1_011n p1_016n p1_017n p1_018n p1_019n p1_027n p1_028n p1_029n p1_030n p1_037n p1_038n p1_039n p1_040n p1_041n p1_042n p1_048n p1_049n p1_050n p1_051n p1_052n p1_053n p1_058n p1_059n p1_060n p1_061n p1_064n p1_065n p1_066n p1_067n p1_069n p1_071n)
Other is all other race groups alone and all multiple race categories that do not include Black or African American (p1_005n p1_006n p1_007n p1_008n p1_012n p1_013n p1_014n p1_015n p1_020n p1_021n p1_022n p1_023n p1_024n p1_025n p1_031n p1_032n p1_033n p1_034n p1_035n p1_036n p1_043n p1_044n p1_045n p1_046n p1_054n p1_055n p1_056n p1_057n p1_062n p1_068n)

Prison addresses were geocoded using the census geocoder at https://geocoding.geo.census.gov/geocoder/

The census ZCTA to census block file (https://www2.census.gov/geo/docs/maps-data/data/rel2020/zcta520/tab20_zcta520_tabblock20_natl.txt was used to assign blockids to the zip codes. 

Variable names in the Kentucky block returning data are:
blockid – Census Block ID code
district_h – 2020 house district
district_s – 2020 senate district
total – Census total population
white – Census white population
black – Census black population
other – Census other population
adjtotal – Population adjusted for returning prisoners to sentencing county
adjwhite - Population adjusted for returning white prisoners to sentencing county
adjblack – Population adjusted for returning black prisoners to sentencing county
adjother - Population adjusted for returning other prisoners to sentencing county

Variable names in the Kentucky district reallocation results file are:
district_h – 2020 house district
district_s – 2020 senate district
total – Census total population
white – Census white population
black – Census black population
other – Census other population
pctwhite – Census percent white
adjtotal – Population adjusted for returning prisoners to sentencing county
adjwhite - Population adjusted for returning white prisoners to sentencing county
adjblack – Population adjusted for returning black prisoners to sentencing county
adjother - Population adjusted for returning other prisoners to sentencing county
changepop – Adjusted population minus unadjusted population
changew – Adjusted white population minus unadjusted white population 
changeb – Adjusted black population minus unadjusted black population 
changeo – Adjusted other population minus unadjusted other population 
toosmall – district is legal too small if reallocation were required
toobig – district is legally too big if reallocation were required

Summary data in the Kentucky district reallocation results are:
Total pop – Census total population
Target size – Census total population divided by the number of districts
5% limit – Target size divided by 5%
Min size – Target size minus the 5 percent limit. Any adjusted districts smaller than the 5% limit are legally too small if reallocation were required. 
Max size – Target size plus the 5 percent limit. Any adjusted districts larger than the 5% limit are legally too big if reallocation were required.
Average impact by race - The average district size change if phantom constituents were reallocated by race/ethnicity. Negative values indicate the average person of that respective race benefits from counting phantom constituents at their correctional facility; positive values indicate the average person of that respective race is harmed by counting phantom constituents as residing in a facility elsewhere.

Average adj% by race – Adjusted percent of population by race and ethnicity

##Additional Notes
The blockid in the Kentucky block returning data set is displayed as scientific notation due to the csv format. This variable should be imported as string into statistical software packages or can be transformed in the csv file by formatting the variable as a custom “000000000000000” variable.

Kentucky provided data that included different total populations in state facilities vs. last known zip code. The raw data included 9,689 more individuals in the last known zip code data than in the prison data due to the dataset including information about state inmates serving in county jails across Kentucky. By statute, a portion of state inmates serve their prison sentences in the county jails. The race by prison data only included state prisons. One prison (Southeast State Correctional Complex, n=336) was excluded from the analysis because it was not built on census day). The zip code file also included 46 discontinued zip codes containing a total of 100 individuals. These 46 zip codes and 100 returnees were dropped from the analysis along with the 176 who were out of state. We calculated two sets of analyses in this case. The first used the raw data provided by the state and thus the total populations are not identical. The adjusted state population includes 9,747 more residents. We also adjusted the total population returned to different zip codes by keeping the returned equal to the number subtracted but returning the original proportion to the counties. For example, if a state gave us a total prison population of 1000 and a county of origin population of 1500, we changed the returning population to 1000, so each county would receive 2/3rd of what is listed in the county of origin file. These are the “balanced” results. The adjusted balanced results include 50 fewer residents due to those living in discontinued zip codes being dropped after the populations were balanced. We provide the balanced results, unbalanced results are available upon request. Balancing changed the results at the margins for Kentucky. There are two senate seats (17 and 22) that became unconstitutionally too big and one (36) that is no longer unconstitutionally too small.  
The adjusted data may include blocks with negative numbers. This is due to a combination of factors. First, the Census’ use of “differential privacy” introduces random noise into the populations of census blocks. We cannot directly address this problem, though we note it is random and thus should, if anything, provide a more conservative estimate of the impact of prison gerrymandering than raw data would. 

Second, in many states, we did not receive the exact same data as the Census (we requested data for April 1, 2020, the day the Census gets data from every state), and this may lead to some differences in incarcerated populations. 

Less frequently, some facility addresses do not match between the Census’ group quarters populations and detailed data. Where possible, we adjusted the location to match the Census group quarters data, but in some situations, there was no population local to the facility listed in Group Quarters in the census’ data. In these cases, we have confirmed that the Census bureau does identify a facility as being in the location we identify using census block maps available at the census website (https://www.census.gov/geographies/reference-maps/2020/geo/2020-census-block-maps.html). Bell County Forestry Camp is located in located in block 210139611001005 even though the Census group quarters population is 0. All negative adjusted block populations are due to identifying a prison in the block with a population larger than the Census population.  

This file and its associated README were created by Rory Kramer, Denise Wilson and Brianna Remster of Villanova University.