2016 Election Campaign Contributions¶
In [1]:
# For data
import pandas as pd
from pandas import Series,DataFrame
import numpy as np
# For visualization
import matplotlib.pyplot as plt
import seaborn as sns
sns.set_style('darkgrid')
%matplotlib inline
from __future__ import division
import datetime
#use to get info from web
import requests
#stringIO to work with csv
from StringIO import StringIO
In [2]:
doner_df = pd.read_csv('Election_Doner_Data.csv', index_col=False, dtype={
'cmte_id': object, 'cand_id': object, 'cand_nm': object, 'contbr_nm': object, 'contbr_city': object, 'contbr_st': object,
'contbr_zip': object, 'contbr_employer': object, 'contbr_occupation': object, 'contb_receipt_amt': float,
'contb_receipt_dt': object, 'receipt_desc': object,'memo_cd': object, 'memo_text': object, 'form_tp': object,
'file_num': int, 'tran_id': object, 'election_tp': object})
In [3]:
doner_df.info()
In [4]:
doner_df.head()
Out[4]:
As we can see there are 7,440,252 rows 18 columns and some null values
This gives us some useful info like the canidates name, contributors name, state, zip, eomployer, occupation, ammount and date
In [5]:
doner_df['contb_receipt_amt'].value_counts()
Out[5]:
The top donation ammount was $25 and donations tend to stick around even values, 5, 10, 25, 50, 100
We can also see that there are some negitive values, this is because the FEC also tracks Refunds, these can be Excessive Contributions, Illegal Contributions or Prohibited Contributions we don't need those because we only want too see how much a canidate raises
In [6]:
# removing negitives from our main dataframe
doner_df = doner_df[doner_df.contb_receipt_amt > 0]
Next we will clean up the data a tad by adding party affiliation¶
In [7]:
# Get a list of all canidates
canidates = doner_df.cand_nm.unique()
canidates
Out[7]:
In [8]:
# dictionary of party affiliation
party_map = {'Rubio, Marco': 'Republican',
'Santorum, Richard J.': 'Republican',
'Perry, James R. (Rick)': 'Republican',
'Carson, Benjamin S.': 'Republican',
"Cruz, Rafael Edward 'Ted'": 'Republican',
'Paul, Rand': 'Republican',
'Clinton, Hillary Rodham': 'Democrat',
'Sanders, Bernard': 'Democrat',
'Fiorina, Carly': 'Republican',
'Huckabee, Mike': 'Republican',
'Pataki, George E.': 'Republican',
"O'Malley, Martin Joseph": 'Democrat',
'Graham, Lindsey O.': 'Republican',
'Bush, Jeb': 'Republican',
'Trump, Donald J.': 'Republican',
'Jindal, Bobby': 'Republican',
'Christie, Christopher J.': 'Republican',
'Walker, Scott': 'Republican',
'Stein, Jill': '3rd Party',
'Webb, James Henry Jr.': 'Democrat',
'Kasich, John R.': 'Republican',
'Gilmore, James S III': 'Republican',
'Lessig, Lawrence': 'Democrat',
'Johnson, Gary': '3rd Party',
'McMullin, Evan': '3rd Party'}
# creating a party column and mapping party to canidate
doner_df['Party'] = doner_df.cand_nm.map(party_map)
In [9]:
doner_df.head(1)
Out[9]:
In [10]:
# A count of how many donations were to each campaign
doner_df.groupby("cand_nm")["contb_receipt_amt"].count().sort_values()
Out[10]:
In [11]:
# The Total the canidate raised
cand_amount = doner_df.groupby('cand_nm')['contb_receipt_amt'].sum().sort_values()
i = 0
for don in cand_amount:
print '%s raised %.0f' %(cand_amount.index[i],don)
i += 1
In [12]:
cand_amount.plot(kind='bar',figsize=(14,6))
Out[12]:
In [13]:
# Average
avg_donations = doner_df.groupby("cand_nm").mean().sort_values(by="contb_receipt_amt")
avg_donations["contb_receipt_amt"].plot(kind="bar",figsize=(12,5))
Out[13]:
Thats a bit easier to see and understand just how much Hillary out funded everyone¶
In [14]:
doner_df.groupby('Party')['contb_receipt_amt'].sum().sort_values().plot(kind='bar')
Out[14]:
While there were a ton of Republican canidates the Democrats specifically Hillary significantly out raised them in total¶
Lets look at donations by Careers¶
In [15]:
# Pivot table much like in excel
job_pt = doner_df.pivot_table('contb_receipt_amt', index='contbr_occupation', columns = 'Party', aggfunc='sum')
In [16]:
job_pt["Total"] = job_pt.sum(axis=1)
job_pt.head()
Out[16]:
In [17]:
job_pt = job_pt.sort_values(by="Total",ascending=True)
In [18]:
job_pt.info()
Okay so this has 127,268 'jobs' but includes blanks errors and duplicates so we clearly dont want all
Lets try grouping by ammount donated and see if that gets us some cleaner information lets try 2 million
In [19]:
job_pt = job_pt[job_pt['Total'] > 2000000]
In [20]:
# 44 jobs are over $2,000,000
job_pt.shape
Out[20]:
In [21]:
job_pt
Out[21]:
In [22]:
# remove these rows
job_pt.drop(['INFORMATION REQUESTED PER BEST EFFORTS', 'INFORMATION REQUESTED', 'RETIRED',
'CONDUIT TOTAL LISTED IN AGG. FIELD'],axis=0,inplace=True)
#combine like rows
job_pt.loc['SELF-EMPLOYED'] = job_pt.loc['SELF-EMPLOYED'] + job_pt.loc['SELF EMPLOYED']
#remove self employed without dash
job_pt.drop('SELF EMPLOYED',inplace=True)
In [23]:
job_pt[['Democrat','Republican']].plot(kind='barh',figsize=(10,16),cmap='bwr')
Out[23]:
Here you can see the breakdown by job and how they donate differently this looks at sum
but how about if we look at how many people donated instead of the ammount they did
In [24]:
# Pivot table much like in excel
job_pt = doner_df.pivot_table('contb_receipt_amt', index='contbr_occupation', columns = 'Party', aggfunc='count')
In [25]:
job_pt["Total"] = job_pt.sum(axis=1)
job_pt.head()
Out[25]:
In [26]:
job_pt = job_pt.sort_values(by="Total",ascending=True)
In [27]:
job_pt = job_pt[job_pt['Total'] > 15000]
In [28]:
job_pt.shape
Out[28]:
In [29]:
job_pt.drop(['DISABLED', 'INFORMATION REQUESTED PER BEST EFFORTS', 'RETIRED', 'INFORMATION REQUESTED',
'CONDUIT TOTAL LISTED IN AGG. FIELD'],axis=0,inplace=True)
job_pt.loc['SELF-EMPLOYED'] = job_pt.loc['SELF-EMPLOYED'] + job_pt.loc['SELF EMPLOYED']
job_pt.loc['REGISTERED NURSE'] = job_pt.loc['REGISTERED NURSE'] + job_pt.loc['RN']
job_pt.loc['BUSINESS OWNER'] = job_pt.loc['BUSINESS OWNER'] + job_pt.loc['OWNER']
job_pt.drop('SELF EMPLOYED', inplace=True)
job_pt.drop('RN', inplace=True)
job_pt.drop('OWNER', inplace=True)
In [30]:
job_pt[['Democrat','Republican']].plot(kind='barh',figsize=(10,16),cmap='bwr')
Out[30]:
Here you can see there was a lot more Not employed then there was Attornies
but if you look at the previous graph of sum donated Attornies have a bit more spending money to give away
Looking At Employer¶
In [31]:
employer_df = doner_df.pivot_table('contb_receipt_amt',index='contbr_employer',columns = 'Party',aggfunc='sum')
In [32]:
employer_df.shape
Out[32]:
In [33]:
# once again graphing 398,017 different companies will not only look horrible, but take forever
employer_df = employer_df[employer_df.sum(1) > 350000]
In [34]:
employer_df.shape
Out[34]:
In [35]:
# Thats better
employer_df
Out[35]:
In [36]:
employer_df.drop(['INFORMATION REQUESTED PER BEST EFFORTS', 'INFORMATION REQUESTED', 'RETIRED',
'TWICE REQUESTED NOT YET RECEIVED'],axis=0,inplace=True)
# combine like rows
employer_df.loc['SELF-EMPLOYED'] = employer_df.loc['SELF-EMPLOYED'] + employer_df.loc['SELF EMPLOYED'] + employer_df.loc['SELF']
employer_df.loc['NOT EMPLOYED'] = employer_df.loc['NOT EMPLOYED'] + employer_df.loc['NONE']
# remove self employed without dash
employer_df.drop('SELF EMPLOYED',inplace=True)
employer_df.drop('SELF',inplace=True)
employer_df.drop('NONE',inplace=True)
# These were far to large to allow us to see anything else
employer_df.drop('HOMEMAKER',inplace=True)
employer_df.drop('NOT EMPLOYED',inplace=True)
employer_df.drop('SELF-EMPLOYED',inplace=True)
In [37]:
employer_df["Total"] = employer_df.sum(axis=1)
employer_df = employer_df.sort_values(by="Total",ascending=True)
employer_df.head()
Out[37]:
In [38]:
employer_df[['Democrat','Republican']].plot(kind='barh',figsize=(10,16),cmap='bwr')
Out[38]:
Lets move on to States¶
In [39]:
state_df = doner_df.pivot_table('contb_receipt_amt',index='contbr_st',columns = 'Party',aggfunc='sum')
In [40]:
state_df.head()
Out[40]:
In [41]:
state_df.drop(['20','30','FF','FR','GU','IS','LO','MP','ON','PR','SI','VI','XX','AS','AU','BC','BR','EN','JA',
'LE','SW','U*','UK','AA','BU','PU','SA','TE','ZZ','AB','BA','BE','CH','DU','E','EU','FM','GR','HE','KE','LI','MB',
'NB','NL','NO','NS','OS','QC','SH','SK','SO','ST','TO','W.','GE','N.','C','PW','AE','AM','AP','QB'],inplace=True)
In [42]:
state_df["Total"] = state_df.sum(axis=1)
state_df = state_df.sort_values(by="Total",ascending=True)
state_df.head()
Out[42]:
In [43]:
state_df[['Democrat','Republican']].plot(kind='barh',figsize=(15,16),cmap='bwr')
Out[43]:
Cleaning Data for Rhode Island¶
In [44]:
RI_df = doner_df[doner_df.contbr_st == 'RI']
In [45]:
RI_df.head()
Out[45]:
In [46]:
RI_df.shape
Out[46]:
In [47]:
RI_df.contbr_city.unique()
Out[47]:
Well.... aparently there is 140 "cities" in RI like previously we have to remove data we don't need like those who live in 'CLINTON' RI
Aswell as combine like values, like the man who lives in SMITHFIELD and those who live in 'SMITHFIELDSMITHFIELD'
but first... lets see how many live in each of these 140 cities
In [48]:
RI_df['contbr_city'].value_counts()
Out[48]:
Okay so a lot of the typos/wrong places are just 1-3 people, we could just remove them and it woulnd't effect our data much, but instead lets add them
I found ~23 that were not actual places in RI at all
I want to convert the smaller towns into whatever big town/city they are part of
In [49]:
RI_df = RI_df[~RI_df['contbr_city'].isin(['CHAMBERSBURG', 'CLINTON', 'FAIR HAVEN', 'CLAREMORE', 'FTWORTH', 'HERMITAGE', 'HONOMU',
'KIAWAH ISLAND', 'LAKE FOREST', 'MADISON', 'MIDDLEFIELD', 'PARIS', 'PERU', 'PITTSBURGH', 'POOLESVILLE', 'PORT ST LUCIE',
'POTTSTOWN', 'PREONZO', 'SALT LAKE CITY', 'SCOTTSDALE', 'SPARTANBURG', 'SPEARFISH', 'VACAVILLE', 'INFO REQUESTED',
'REDMOND'])]
In [50]:
RI_df = RI_df.replace(['BARRINGTION'],'BARRINGTON')
RI_df = RI_df.replace(['GLENDALE','HARRISVILLE','MAPLEVILLE','OAKLAND','PASCOAG'],'BURRILLVILLE')
RI_df = RI_df.replace(['CAROLINA','CHARLESTON','CHARLESTOWN'],'CHARLESTOWN')
RI_df = RI_df.replace(['GREENE'],'COVENTRY')
RI_df = RI_df.replace(['E GREENWICH','E. GREENWICH','EAST GREENWICH, RI 02818'],'EAST GREENWICH')
RI_df = RI_df.replace(['E PROVIDENCE','EAST PROVIDENCE','ESAT PROVIDENCE','RIVERSIDE','RUMFORD'],'EAST PROVIDENCE')
RI_df = RI_df.replace(['CLAYVILLE'],'FOSTER')
RI_df = RI_df.replace(['CHAPACHET','CHEPACHET','HARMONY'],'GLOCESTER')
RI_df = RI_df.replace(['ASHAWAY','HOPE VALLEY','ROCKVILLE'],'HOPKINTON')
RI_df = RI_df.replace(['ALBION','FAIR LAWN','LINCON','MANVILLE'],'LINCOLN')
RI_df = RI_df.replace(['ADAMSVILLE','L. COMPTON','LITTLE COMPTO','MANVILLE'],'LITTLE COMPTON')
RI_df = RI_df.replace(['MIDDLETWON'],'MIDDLETOWN')
RI_df = RI_df.replace(['NARRAGANSETT','NARRAGONSETT','SAUNDERSTOWN','SAUNDERSTROWN','NARRAGANSET'],'NARRAGANSETT')
RI_df = RI_df.replace(['BLOCK ISLAND'],'NEW SHOREHAM')
RI_df = RI_df.replace(['N KINGSTOWN','N. KINGSTOWN','NO.KINGSTOWN','NORTH KING','SLOCUM','WCKFORD','WICKFORD'],'NORTH KINGSTOWN')
RI_df = RI_df.replace(['N PROVIDENCE','N. PROVIDENCE','N.PROV.','NO. PROVIDENCE','NORTHPROVIDENCE'],'NORTH PROVIDENCE')
RI_df = RI_df.replace(['FORESTDALE','N SMITHFIELD','SLATERSVILLE'],'NORTH SMITHFIELD')
RI_df = RI_df.replace(['PRUDENCE ISL'],'PORTSMOUTH')
RI_df = RI_df.replace(['SHANNOCK','WOOD RIVER JUNCTION','WYOMING'],'RICHMOND')
RI_df = RI_df.replace(['HOPE','N SCITUATE','N. SCITUATE','N.SCITUATE','NORTH SCITUATE'],'SCITUATE')
RI_df = RI_df.replace(['GREENVILLE','SMITHFIELDSMITHFIELD'],'SMITHFIELD')
RI_df = RI_df.replace(['KINGSTON','PEACE DALE','S KINGSTOWN','S. KINGSTOWN','SOUTH KINGSTOWN','WAKEFIELD','WEDT KINGSTON','WEST KINGSTON'],'SOUTH KINGSTOWN')
RI_df = RI_df.replace(['WARIWCK','WAWICK','WARWICK. RHODE ISLAND'],'WARWICK')
RI_df = RI_df.replace(['W GREENWICH','WSET GREENWICH'],'WEST GREENWICH')
RI_df = RI_df.replace(['WEST WARWWICK'],'WEST WARWICK')
RI_df = RI_df.replace(['BRADFORD'],'WESTERLY')
In [51]:
RI_df['contbr_city'].value_counts()
Out[51]:
Okay so now we got the data into major cities/towns next lets add counties¶
In [52]:
# dictionary of RI Counties
county_map = {'BARRINGTON': 'BRISTOL',
'BRISTOL': 'BRISTOL',
'BURRILLVILLE': 'PROVIDENCE',
'CENTRAL FALLS': 'PROVIDENCE',
'CHARLESTOWN': 'WASHINGTON',
'COVENTRY': 'KENT',
'CRANSTON': 'PROVIDENCE',
'CUMBERLAND': 'PROVIDENCE',
'EAST GREENWICH': 'KENT',
'EAST PROVIDENCE': 'PROVIDENCE',
'EXETER': 'WASHINGTON',
'FOSTER': 'PROVIDENCE',
'GLOCESTER': 'PROVIDENCE',
'HOPKINTON': 'WASHINGTON',
'JAMESTOWN': 'NEWPORT',
'JOHNSTON': 'PROVIDENCE',
'LINCOLN': 'PROVIDENCE',
'LITTLE COMPTON': 'NEWPORT',
'MIDDLETOWN': 'NEWPORT',
'NARRAGANSETT': 'WASHINGTON',
'NEWPORT': 'NEWPORT',
'NEW SHOREHAM': 'WASHINGTON',
'NORTH KINGSTOWN': 'WASHINGTON',
'NORTH PROVIDENCE': 'PROVIDENCE',
'NORTH SMITHFIELD': 'PROVIDENCE',
'PAWTUCKET': 'PROVIDENCE',
'PORTSMOUTH': 'NEWPORT',
'PROVIDENCE': 'PROVIDENCE',
'RICHMOND': 'WASHINGTON',
'SCITUATE': 'PROVIDENCE',
'SMITHFIELD': 'PROVIDENCE',
'SOUTH KINGSTOWN': 'WASHINGTON',
'TIVERTON': 'NEWPORT',
'WARREN': 'BRISTOL',
'WARWICK': 'KENT',
'WESTERLY': 'WASHINGTON',
'WEST GREENWICH': 'KENT',
'WEST WARWICK': 'KENT',
'WOONSOCKET': 'PROVIDENCE'}
# creating a party column and mapping party to canidate
RI_df['County'] = RI_df.contbr_city.map(county_map)
In [53]:
RI_df['County'].value_counts()
Out[53]:
In [54]:
RI_df.to_csv('RI_Pres_Election_2016.csv')