How the Polls Moved During the 2016 Election¶

Contains¶

  • Major Events
  • The Debates
  • FBI Investigation
  • World Events
  • Rallies for Both Canidates
For all events except Rallies, any event date that didnt have data it was moved to the next date (labeled AFTER)
Because Rallies were so frequent only 1 per day is listed, and if there was no data for the date it was left out of the graph

Source: http://elections.huffingtonpost.com/pollster/2016-general-election-trump-vs-clinton
In [26]:
# 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('whitegrid')
%matplotlib inline

from __future__ import division
from datetime import datetime

# use to get info from web
import requests

# stringIO to work with csv
from StringIO import StringIO
In [27]:
# getting the data
url = 'http://elections.huffingtonpost.com/pollster/2016-general-election-trump-vs-clinton.csv'

source = requests.get(url).text
poll_data = StringIO(source)
In [28]:
# creating the dataframe
poll_df = pd.read_csv(poll_data)
In [30]:
poll_df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1522 entries, 0 to 1521
Data columns (total 17 columns):
Pollster                  1522 non-null object
Start Date                1522 non-null object
End Date                  1522 non-null object
Entry Date/Time (ET)      1522 non-null object
Number of Observations    1013 non-null float64
Population                1522 non-null object
Mode                      1522 non-null object
Trump                     1522 non-null float64
Clinton                   1522 non-null float64
Other                     1098 non-null float64
Undecided                 1460 non-null float64
Pollster URL              1522 non-null object
Source URL                1522 non-null object
Partisan                  1522 non-null object
Affiliation               1522 non-null object
Question Text             661 non-null object
Question Iteration        1522 non-null int64
dtypes: float64(5), int64(1), object(11)
memory usage: 202.2+ KB

Now we see there are 1522 Polls¶

In [31]:
# previewing the data
poll_df.head()
Out[31]:
Pollster Start Date End Date Entry Date/Time (ET) Number of Observations Population Mode Trump Clinton Other Undecided Pollster URL Source URL Partisan Affiliation Question Text Question Iteration
0 Insights West 2016-11-04 2016-11-07 2016-11-08T12:16:30Z 940.0 Likely Voters Internet 41.0 45.0 2.0 8.0 http://elections.huffingtonpost.com/pollster/p... http://www.insightswest.com/news/clinton-is-ah... Nonpartisan None As you may know, there will be a presidential ... 1
1 Insights West 2016-11-04 2016-11-07 2016-11-08T12:16:30Z NaN Likely Voters - Democrat Internet 6.0 89.0 0.0 4.0 http://elections.huffingtonpost.com/pollster/p... http://www.insightswest.com/news/clinton-is-ah... Nonpartisan None As you may know, there will be a presidential ... 1
2 Insights West 2016-11-04 2016-11-07 2016-11-08T12:16:30Z NaN Likely Voters - Republican Internet 82.0 7.0 2.0 6.0 http://elections.huffingtonpost.com/pollster/p... http://www.insightswest.com/news/clinton-is-ah... Nonpartisan None As you may know, there will be a presidential ... 1
3 Insights West 2016-11-04 2016-11-07 2016-11-08T12:16:30Z NaN Likely Voters - independent Internet 38.0 43.0 4.0 7.0 http://elections.huffingtonpost.com/pollster/p... http://www.insightswest.com/news/clinton-is-ah... Nonpartisan None As you may know, there will be a presidential ... 1
4 IBD/TIPP 2016-11-04 2016-11-07 2016-11-08T12:10:06Z 1107.0 Likely Voters Live Phone 43.0 41.0 4.0 5.0 http://elections.huffingtonpost.com/pollster/p... http://www.investors.com/politics/ibd-tipp-pre... Nonpartisan None NaN 1
In [32]:
# averaging the data
avg = pd.DataFrame(poll_df.mean())

#droping the info we dont need
avg.drop('Number of Observations',axis=0,inplace=True)
avg.drop('Question Iteration',axis=0,inplace=True)
In [33]:
# getting the standard deviation
std = pd.DataFrame(poll_df.std())

#dropping the same columns we dont need
std.drop('Number of Observations',axis=0,inplace=True)
std.drop('Question Iteration',axis=0,inplace=True)
In [34]:
# combining them into a dataframe
poll_avg = pd.concat([avg,std],axis=1)

#adding header
poll_avg.columns = ['Average','STD']
In [35]:
# viewing the average and standard deviation
poll_avg
Out[35]:
Average STD
Trump 40.643890 23.566390
Clinton 42.733903 25.298731
Other 5.806011 5.009533
Undecided 9.315068 6.253118
In [36]:
# Adding a difference column to show the difference between clinton/trump in every poll
# A POSITIVE value = Leaning Clinton a NEGITIVE value = Leaning Trump
poll_df['Difference'] = (poll_df.Clinton - poll_df.Trump)/100
In [37]:
# grouping all the polls by day and averaging all the data
poll_df = poll_df.groupby(['Start Date'],as_index=False).mean()

poll_df.head()
Out[37]:
Start Date Number of Observations Trump Clinton Other Undecided Question Iteration Difference
0 2015-05-19 1046.00 34.25 48.75 2.5 14.00 1.0 0.1450
1 2015-06-20 420.75 35.00 47.25 NaN 17.75 1.0 0.1225
2 2015-06-21 1005.00 34.00 51.00 3.0 12.00 1.0 0.1700
3 2015-06-26 890.00 36.75 57.00 6.0 0.00 1.0 0.2025
4 2015-07-09 499.25 35.25 49.50 NaN 16.00 1.0 0.1425
In [38]:
# previewing the difference throughout the election
poll_df.plot('Start Date','Difference',figsize=(12,4),marker='',linestyle='-',color='purple')
Out[38]:
<matplotlib.axes._subplots.AxesSubplot at 0xde112b0>

Now that we have an idea of how the polls moved lets start marking some key events that occured¶

Debates¶

Lets start by looking at the 3 debates¶

  • Sept. 26, 2016 - 1st debate
  • Oct. 9, 2016 - 2nd debate
  • Oct. 19, 2016 - 3rd debate
In [39]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(226,262))

#sept 26 1st debate
plt.axvline(x=225+2,linewidth=4,color='grey', alpha=0.5)
#oct 9th 2nd debate
plt.axvline(x=232+8,linewidth=4,color='grey', alpha=0.5)
#oct 19th 3rd debate
plt.axvline(x=232+18,linewidth=4,color='grey', alpha=0.5)
Out[39]:
<matplotlib.lines.Line2D at 0xe1fab38>

Keeping in mind events may not effect polls untill the following few days depending on what time the event happened and what time the poll was done the 3rd debate looks like it may have boosted Clinton 2 days later, but maybe that was something else lets look at some more events¶

Major Election Events By Month¶

Blue = Hillary Event | Red = Trump Event | Grey = Both
For this section if there wasn't a poll on a important date, the following day is marked instead

July 2016¶

  • July 5, 2016 - FBI Announces Its Not Recommending Criminal Charges Against Clinton
  • July 12, 2016 - Sanders Endorses Clinton for President
  • July 16, 2016 - Trump Officially Introduces Mike Pence as His Running Mate
  • July 21, 2016 - Trump Accepts Republican Nomination
  • July 22, 2016 - Clinton Chooses Tim Kaine for Her Running Mate
  • July 29, 2016 - Day AFTER Clinton Makes History Accepting Democratic Nomination
In [40]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(159,182))

plt.axvline(x=162,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=167,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=171,linewidth=4,color='red', alpha=0.5)
plt.axvline(x=174,linewidth=4,color='red', alpha=0.5)
plt.axvline(x=175,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=180,linewidth=4,color='blue', alpha=0.5)
Out[40]:
<matplotlib.lines.Line2D at 0xe4e7400>

August 2016¶

  • Aug. 10, 2016 - Newly Released Emails Highlight Clinton Foundation’s Ties to State Department
  • Aug. 17, 2016 - Donald Trump hands Kellyanne Conway, Stephen Bannon new roles in campaign
In [41]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(183,208))

plt.axvline(x=189,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=196,linewidth=4,color='red', alpha=0.5)
Out[41]:
<matplotlib.lines.Line2D at 0xe336588>

September 2016¶

  • Sept. 10, 2016 - ‘Basket of Deplorables’ comment from Clinton
  • Sept. 11, 2016 - Clinton ‘Overheats’ at 9/11 Memorial, Campaign Reveals Clinton Has Pneumonia
  • Sept. 18, 2016 - Both Canidates react to New York Bombing
  • Sept. 26, 2016 - The First Presidential Debate
In [42]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(209,231))

plt.axvline(x=215,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=216,linewidth=4,color='blue', alpha=0.5)
plt.axvline(x=221,linewidth=4,color='grey', alpha=0.5)
plt.axvline(x=227,linewidth=4,color='grey', alpha=0.5)
Out[42]:
<matplotlib.lines.Line2D at 0xe840d68>

October 2016¶

  • Oct. 7, 2016 - WikiLeaks Release Thousands of Clinton Campaign Chair’s Emails
  • Oct. 7, 2016 - Trump Is Heard Making Lewd Comments About Women in 2005 Video
  • Oct. 9, 2016 - The Second Presidential Debate
  • Oct. 12, 2016 - Women Accuse Trump of Inappropriately touching them
  • Oct. 19, 2016 - The Final Presidential Debate
  • Oct. 28, 2016 - The FBI Reviews New Emails Tied to Clinton Private Email Server Investigation
In [43]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(232,262))

plt.axvline(x=238,linewidth=4,color='grey', alpha=0.5)
plt.axvline(x=238,linewidth=4,color='grey', alpha=0.5)
plt.axvline(x=240,linewidth=4,color='grey', alpha=0.5)
plt.axvline(x=243,linewidth=4,color='red', alpha=0.5)
plt.axvline(x=250,linewidth=4,color='grey', alpha=0.5)
plt.axvline(x=259,linewidth=4,color='blue', alpha=0.5)
Out[43]:
<matplotlib.lines.Line2D at 0xf0e83c8>

Now lets look at the FBI announcing there investigation into Clinton 11 days before the election¶

FBI Clinton Investigation 11 Days Before Election¶

  • Oct. 28th, 2016 - James Comey informs congress they are re opening a probe into Clinton
In [44]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(257,266))

plt.axvline(x=232+27,linewidth=4,color='blue', alpha=0.5)
Out[44]:
<matplotlib.lines.Line2D at 0xe59f908>

Doesn't seem to have had much of an effect, sadly HuffPost Pollster Data stops before the 6th when Comey announced that the new evidence found did not change the agency conclusion that Hillary shouldn't face criminal charges

Lets move on to World events

Major World Events By Month¶

Orange = Terror Attack | Cyan = World Politics | Green = Other¶

June 2016¶

  • June 12, 2016 - Orlando Nightclub Terror Attack
  • June 23, 2016 - Brexit
  • June 28, 2016 - Istanbul Turkey Airport Terror Attack
In [46]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(134,159))

plt.axvline(x=143,linewidth=4,color='orange', alpha=0.5)
plt.axvline(x=152,linewidth=4,color='cyan', alpha=0.5)
plt.axvline(x=157,linewidth=4,color='orange', alpha=0.5)
Out[46]:
<matplotlib.lines.Line2D at 0xf6936a0>

July 2016¶

  • July 14, 2016 - France Nice Terror Attack
  • July 15 2016 - Turkey Coup Attempt
In [47]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(159,182))

plt.axvline(x=169,linewidth=4,color='orange', alpha=0.5)
plt.axvline(x=170,linewidth=4,color='green', alpha=0.5)
Out[47]:
<matplotlib.lines.Line2D at 0xe08d160>

August 2016¶

  • Aug. 5, 2016 - Olympics Start
  • Aug. 21, 2016 - Olympics End
  • Aug. 25, 2016 - Italy Earthquake
In [48]:
poll_df.plot('Start Date','Difference',figsize=(12,6),marker='o',linestyle='-',color='purple',xlim=(183,208))

plt.axvline(x=185,linewidth=4,color='green', alpha=0.5)
plt.axvline(x=200,linewidth=4,color='green', alpha=0.5)
plt.axvline(x=203,linewidth=4,color='green', alpha=0.5)
Out[48]:
<matplotlib.lines.Line2D at 0xf9f2518>

From skimming headlines I didn't find much for Sept. or Oct. So lets move on to Rallies!¶

Trump Rallies¶

Dates that are missing poll data were left out
Only included rallies with citations | Only includes 1 per day

Source: https://en.wikipedia.org/wiki/List_of_rallies_for_the_Donald_Trump_presidential_campaign,_2016

June 2016¶

  • June 1, 2016 - Sacramento, CA
  • June 2, 2016 - San Jose, CA
  • June 3, 2016 - Sacramento, CA
  • June 10, 2016 - Richmond, VA
  • June 11, 2016 - Tampa, FL
  • June 14, 2016 - Greensboro, NC
  • June 15, 2016 - Atlanta, GA
  • June 16, 2016 - Dallas, TX
  • June 17, 2016 - The Woodlands, TX
  • June 18, 2016 - Phoenix, AZ
  • June 28, 2016 - St. Clairsville, OH
  • June 29, 2016 - Bangor, ME
Out[49]:
<matplotlib.lines.Line2D at 0xfc41080>

July 2016¶

  • July 5, 2016 - Raleigh, NC
  • July 6, 2016 - Cincinnati, OH
  • July 12, 2016 - Westfield, IN
  • July 25, 2016 - Winston-Salem, NC
  • July 27, 2016 - Toledo, OH
  • July 29, 2016 - Denver, CO
Out[50]:
<matplotlib.lines.Line2D at 0xfe03438>

August 2016¶

  • Aug. 1, 2016 - Mechanicsburg, PA
  • Aug. 4, 2016 - Portland, ME
  • Aug. 5, 2016 - Green Bay, WA
  • Aug. 6, 2016 - Windham, NH
  • Aug. 9, 2016 - Wilmington, NC
  • Aug. 10, 2016 - Sunrise, FL
  • Aug. 11, 2016 - Kissimmee, FL
  • Aug. 12, 2016 - Erie, PA
  • Aug. 13, 2016 - Fairfield, CT
  • Aug. 16, 2016 - West Bend, WI
  • Aug. 18, 2016 - Charlotte, NC
  • Aug. 19, 2016 - Dimondale, MI
  • Aug. 20, 2016 - Fredericksburg, VA
  • Aug. 22, 2016 - Akron, OH
  • Aug. 23, 2016 - Austin, TX
  • Aug. 24, 2016 - Jackson, MS
  • Aug. 25, 2016 - Manchester, NH
  • Aug. 31, 2016 - Phoenix, AZ
Out[51]:
<matplotlib.lines.Line2D at 0x10dc5898>

September 2016¶

  • Sept. 1, 2016 - Wilmington, OH
  • Sept. 6, 2016 - Greenville, NC
  • Sept. 9, 2016 - Pensacola, FL
  • Sept. 12, 2016 - Asheville, NC
  • Sept. 15, 2016 - Laconia, NH
  • Sept. 16, 2016 - Miami, FL
  • Sept. 17, 2016 - Colorado Springs, CO
  • Sept. 19, 2016 - Estero, FL
  • Sept. 20, 2016 - Kenansville, NC
  • Sept. 21, 2016 - Toledo, OH
  • Sept. 22, 2016 - Chester Township, PA
  • Sept. 27, 2016 - Melbourne, FL
  • Sept. 28, 2016 - Waukesha, WI
  • Sept. 29, 2016 - Bedford, NH
  • Sept. 30, 2016 - Novi, MI
Out[52]:
<matplotlib.lines.Line2D at 0x110d5710>

October 2016¶

  • Oct. 1, 2016 - Manheim, PA
  • Oct. 3, 2016 - Loveland, CO
  • Oct. 4, 2016 - Prescott Valley, AZ
  • Oct. 5, 2016 - Henderson, NV
  • Oct. 10, 2016 - Ambridge, PA
  • Oct. 11, 2016 - Panama City Beach, FL
  • Oct. 12, 2016 - Lakeland, FL
  • Oct. 13, 2016 - West Palm Beach, FL
  • Oct. 14, 2016 - Charlotte, NC
  • Oct. 15, 2016 - Portsmouth, NH
  • Oct. 20, 2016 - Delaware, OH
  • Oct. 21, 2016 - Newtown Township, PA
  • Oct. 22, 2016 - Virginia Beach, VA
  • Oct. 25, 2016 - Sanford, FL
  • Oct. 26, 2016 - Kinston, NC
  • Oct. 27, 2016 - Geneva, OH
  • Oct. 28, 2016 - Cedar Rapids, IA
  • Oct. 29, 2016 - Phoenix, AZ
  • Oct. 30, 2016 - Albuquerque, NM
  • Oct. 31, 2016 - Warren, MI
Out[55]:
<matplotlib.lines.Line2D at 0x111926d8>

All Trump Rallies¶

Out[63]:
<matplotlib.lines.Line2D at 0x1546d4e0>

Clinton Rallies¶

Source: https://hillaryspeeches.com/scheduled-events/¶

June 2016¶

  • June 1, 2016 - Newark, NJ
  • June 2, 2016 - El Centro, CA
  • June 4, 2016 - Fresno, CA
  • June 5, 2016 - Sacramento, CA
  • June 6, 2016 - Lynwood, CA
  • June 14, 2016 - Pittsburgh, PA
  • June 21, 2016 - Columbus, OH
  • June 27, 2016 - Cincinnati, OH
Out[64]:
<matplotlib.lines.Line2D at 0x15644a58>

July 2016¶

  • July 5, 2016 - Charlotte, NC
  • July 12, 2016 - Portsmouth, NH
  • July 13, 2016 - Springfield, IL
  • July 14, 2016 - Annandale, VA
  • July 18, 2016 - Cincinnati, OH
  • July 22, 2016 - Tampa, FL
  • July 23, 2016 - Miami, FL
  • July 29, 2016 - Harrisburg, PA
  • July 30, 2016 - Youngstown, OH
  • July 31, 2016 - Columbus, OH
Out[65]:
<matplotlib.lines.Line2D at 0x1555e748>

August 2016¶

  • Aug. 1, 2016 - Omaha, NE
  • Aug. 4, 2016 - Las Vegas, NV
  • Aug. 8, 2016 - St. Petersburg, FL
  • Aug. 10, 2016 - Des Moines, IA
  • Aug. 15, 2016 - Scranton, PA
  • Aug. 16, 2016 - Philadelphia, PA
  • Aug. 17, 2016 - Cleveland, OH
  • Aug. 25, 2016 - Reno, NV
Out[66]:
<matplotlib.lines.Line2D at 0x15ae2cf8>

September 2016¶

  • Sept. 5, 2016 - Cleveland, OH
  • Sept. 6, 2016 - Tampa, FL
  • Sept. 8, 2016 - Charlotte, NC
  • Sept. 15, 2016 - Greensboro, NC
  • Sept. 19, 2016 - Philadelphia, PA
  • Sept. 21, 2016 - Orlando, FL
  • Sept. 27, 2016 - Raleigh, NC
  • Sept. 28, 2016 - Durham, NH
  • Sept. 30, 2016 - Fort Pierce, FL
Out[67]:
<matplotlib.lines.Line2D at 0x15fda160>

October 2016¶

  • Oct. 3, 2016 - Toledo, OH
  • Oct. 4, 2016 - Harrisburg, PA
  • Oct. 10, 2016 - Columbus, OH
  • Oct. 11, 2016 - Miami, FL
  • Oct. 12, 2016 - Pueblo, CO
  • Oct. 21, 2016 - Cleveland, OH
  • Oct. 22, 2016 - Philadelphia, PA
  • Oct. 23, 2016 - Raleigh, NC
  • Oct. 24, 2016 - Manchester, NH
  • Oct. 25, 2016 - Coconut Creek, FL
  • Oct. 26, 2016 - Tampa, FL
  • Oct. 27, 2016 - Winston Salem, NC
  • Oct. 28, 2016 - Des Moines, IA
  • Oct. 30, 2016 - Wilton Manors, FL
  • Oct. 31, 2016 - Cincinnati, OH
Out[68]:
<matplotlib.lines.Line2D at 0x161bd7b8>

All Clinton Rallies¶

Out[71]:
<matplotlib.lines.Line2D at 0x1819ec88>

Trump vs. Clinton Rallies¶

Trump = Red | Clinton = Blue | Purple = Both¶

Out[72]:
<matplotlib.lines.Line2D at 0x188baa20>