The primary dataset can be found in /data/input/election_county2016.csv
See Codebook in /data/input/codebook_election.xlsx for variable definitions.
For 2016 and 2012 election variables: Presidential results were obtained from the 3 election sources linked in the codebook as well as MS SoS website. Data accuracy was checked against mid-November 2016 preliminary results orginially hosted on Townhall.com (but not included). 3rd party breakdown (Johnson, Stein, Other) is missing for Kansas and Maine. Data for the Kansas senate race are similarily missing.
For constructed variables (e.g. racial diversity), descriptive definitions are in the Codebook. In general, enough detail is provided to allow for reproduduction.
Specifically for the 2012 Nytrank variable, the code to generate:
## Nytrank http://nyti.ms/2k1WWFM
### Rank Vars
library("data.table")
dt <- dt[!is.na(unemp), u3rank := rank(unemp,ties.method = "min")]
dt <- dt[!is.na(ed_degree), edrank := rank(-ed_degree, ties.method = "min")]
dt <- dt[!is.na(disability), disrank := rank(disability, ties.method = "min")]
dt <- dt[!is.na(le10), lerank := rank(-le10, ties.method = "min")]
dt <- dt[!is.na(hh_income12), mhirank := rank(-hh_income12, ties.method = "min")]
dt <- dt[!is.na(obese11), obeserank := rank(obese, ties.method = "min")]
### Replace NAs
dt$u3rank[is.na(dt$u3rank)] <- 0
#dt$edrank[is.na(dt$edrank)] <- 0
#dt$disrank[is.na(dt$disrank)] <- 0
#dt$lerank[is.na(dt$lerank)] <- 0
dt$mhirank[is.na(dt$mhirank)] <- 0
#dt$obeserank[is.na(dt$obeserank)] <- 0
## Combine
dt$nytrank <- (dt$edrank + dt$u3rank + dt$disrank
+ dt$lerank + dt$mhirank + dt$obeserank)/
((as.numeric((dt$edrank!=0))) + (as.numeric((dt$u3rank!=0))) +
(as.numeric((dt$disrank!=0))) + (as.numeric((dt$lerank!=0))) +
(as.numeric((dt$mhirank!=0))) + (as.numeric((dt$obeserank!=0)))
)
In addition to the codebook, two analysis files are provided to provide instructive examples of using the dataset. A few rough plots of the data using mapinseconds.com are also provided.