SuperPAC Receipts by State
A colleague exposed me to the R enviroment. It smoted me and I looked for non-work datasets. I found The Sunlight Foundation and started digging through their data, and I found their reports on Super PAC spending. I downloaded the Super PAC Summary data and broke out R.
I wanted to break out the Independent Expenditures by State, and map them. I read in the data and tidied it up a bit.
suppressPackageStartupMessages(require(mapproj))
suppressPackageStartupMessages(require(ggplot2))
suppressPackageStartupMessages(require(grid))
suppressPackageStartupMessages(library(maps))
suppressPackageStartupMessages(require(plyr))
pac <- read.csv("./committee_summary.csv", header = TRUE)
names(pac)[12] <- "receipts"
str(pac)## 'data.frame': 1027 obs. of 20 variables:
## $ Name : Factor w/ 1025 levels "1911 UNITED",..: 782 56 681 525 428 380 997 210 815 334 ...
## $ Committee.ID : Factor w/ 1027 levels "C00235929","C00363390",..: 62 30 100 23 98 120 175 31 534 60 ...
## $ Treasurer : Factor w/ 449 levels "","AARON NOSBISCH",..: 397 89 395 250 60 374 49 71 128 437 ...
## $ Street_1 : Factor w/ 450 levels "","1 PENN PLAZA",..: 258 73 28 283 282 214 143 131 113 264 ...
## $ Street_2 : Factor w/ 40 levels "","#211","148 NASSAU AVE UPPER",..: 26 29 18 38 1 1 1 1 1 1 ...
## $ City : Factor w/ 241 levels "","AKRON","ALBANY",..: 228 228 228 228 228 228 113 228 228 214 ...
## $ ZIP.code : int 20004 20005 20005 20005 20005 20001 30043 20036 20036 33606 ...
## $ state : Factor w/ 51 levels "","AK","AL","AR",..: 9 9 9 9 9 9 12 9 9 11 ...
## $ Connected.Org.Name : Factor w/ 28 levels "","9-9-9 FUND",..: 1 16 28 16 16 1 1 1 22 1 ...
## $ Political.Orientation : Factor w/ 3 levels "Backs Democrats",..: 2 2 1 1 1 2 2 2 1 2 ...
## $ Filing.frequency : Factor w/ 3 levels "Monthly","Quarterly",..: 1 1 1 2 2 1 1 1 2 1 ...
## $ receipts : num 1.32e+08 8.01e+07 6.39e+07 3.49e+07 2.77e+07 ...
## $ Total.unitemized.contributions: num 298 188926 279650 43230 243088 ...
## $ cash.on.hand : num 24203768 6423726 10086434 7461199 5196881 ...
## $ last.report.date : Factor w/ 29 levels "","10/11/12",..: 4 4 4 4 4 4 4 4 4 4 ...
## $ total.IEs : num 1.43e+08 1.05e+08 6.62e+07 3.75e+07 3.08e+07 ...
## $ IEs.support.dems : num 0 0 0 3609417 865049 ...
## $ IEs.oppose.dems : num 88572359 95858190 0 0 0 ...
## $ IEs.support.reps : num 14090892 8477066 0 0 0 ...
## $ IEs.oppose.reps : num 39992095 375217 66182181 33868125 29709982 ...PAC is a data frame with many columns. For this purpose, I’m interested in the receipts and the state identifiers.
I needed to summarize the receipts and ensure that the states matched the codes from the states_map object. And then I had to override the “district of columbia”. At that point, I could plot out a quick view.
pdata <- ddply(pac, "state", summarize, Sum = sum(receipts, na.rm = TRUE))
states_map <- map_data("state")
pdata$statelc <- tolower(state.name[match(pdata$state, state.abb)])
pdata$statelc[9] <- "district of columbia"
ggplot(data = pdata, aes(x = state, y = Sum)) + geom_bar(width = 1, stat = "identity") +
theme(legend.position = "none") + labs(x = "State", y = "Receipts") + ggtitle("SuperPAC Receipts by State Comparison")
Unsuprisingly, most SuperPAC money flows to PACs headquartered in Washington, DC. So much money that DC will dominate subsequent plots; I decided to get rid of DC.
pdata.noDC = subset(pdata, state != "DC")
ggplot(pdata.noDC, aes(x = state, y = Sum)) + geom_bar(width = 1, stat = "identity") +
theme(legend.position = "none") + labs(x = "State", y = "Receipts") + ggtitle("SuperPAC Receipts by State Comparison\nDC omitted")
Much better.
A quick coloring of the states with a gradient showed
ggplot(pdata.noDC, aes(map_id = statelc, fill = Sum)) + geom_map(map = states_map) +
expand_limits(x = states_map$long, y = states_map$lat) + coord_map("polyconic")
that the gradient treatment doesn’t really help distinguish between different levels: is Washington more or less than Oregon? But note that Nebraska isn’t participating in Super PAC spending at all.
Quantization provided a better view.
qs = quantile(pdata.noDC$Sum, c(0, 0.2, 0.4, 0.6, 0.8, 1))
qsf = sprintf("$%.2f", qs)
qlabels = paste(head(qsf, -1), tail(qsf, -1), sep = " - ")
pdata.noDC$Sum_q <- cut(pdata.noDC$Sum, qs, labels = qlabels, include.lowest = TRUE)
pal <- colorRampPalette(c("grey80", "darkgreen"))(5)
clean_theme <- theme(axis.title = element_blank(), axis.text = element_blank(),
panel.background = element_blank(), panel.grid = element_blank(), axis.ticks.length = unit(0,
"cm"), complete = TRUE)
ggplot(pdata.noDC, aes(map_id = statelc, fill = Sum_q)) + geom_map(map = states_map,
colour = "black") + scale_fill_manual(values = pal) + expand_limits(x = states_map$long,
y = states_map$lat) + coord_map("polyconic") + labs(fill = "Receipts", x = "Longitude",
y = "Latitude") + ggtitle("SuperPAC Receipts by State\nof Registration\nDC Omitted") +
clean_theme
Now, we can see that Super PACs in Oregon take in fewer receipts than those in Washington.
This was a quick way to take state-coded data and present it graphically. I would have been “done”, except that I noticed that the data has columns such as “IEs.support.dems” and “IEs.support.reps”. Sunlight tells me that this spending is either positive (support) or negative (oppose). Doing something with that information seemed interesting, and is covered in the next post.