All Projects → Prooffreader → Election_2016_data

Prooffreader / Election_2016_data

Scraped data from the 2016 U.S. Election (President, Senate, House, Governor) and primaries, ballot measures and exit polls

Projects that are alternatives of or similar to Election 2016 data

Yolov3 Point
从零开始学习YOLOv3教程解读代码+注意力模块(SE,SPP,RFB etc)
Stars: ✭ 119 (-1.65%)
Mutual labels:  jupyter-notebook
Keywords2vec
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Models
Model zoo for genomics
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Chatbot Retrieval
Dual LSTM Encoder for Dialog Response Generation
Stars: ✭ 1,547 (+1178.51%)
Mutual labels:  jupyter-notebook
Pytorch Rl
Tutorials for reinforcement learning in PyTorch and Gym by implementing a few of the popular algorithms. [IN PROGRESS]
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Time Series Classification And Clustering With Reservoir Computing
Library for implementing reservoir computing models (echo state networks) for multivariate time series classification and clustering.
Stars: ✭ 120 (-0.83%)
Mutual labels:  jupyter-notebook
Bostonml
Stars: ✭ 120 (-0.83%)
Mutual labels:  jupyter-notebook
Bobsql
demos, scripts, samples, and code from the two bobs who work at Microsoft on SQL Server
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Machine learning model
机器学习基本模型算法介绍(附加案例)
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Batchai
Repo for publishing code Samples and CLI samples for BatchAI service
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Pandas Videos
Jupyter notebook and datasets from the pandas Q&A video series
Stars: ✭ 1,716 (+1318.18%)
Mutual labels:  jupyter-notebook
Western constellations atlas of space
Code, data, and instructions to map every star you can see from Earth
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Multilstm
keras attentional bi-LSTM-CRF for Joint NLU (slot-filling and intent detection) with ATIS
Stars: ✭ 122 (+0.83%)
Mutual labels:  jupyter-notebook
Naturallanguagerecommendations
Getting recommendations from natural language
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Eeg Classification
This project was a joint effort with the neurology labs at UNL and UCD Anschutz to use deep learning to classify EEG data.
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Limperg python
Repository with material for the Limperg Python course by Ties de Kok.
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Deep learning explorations
Codes and experiments while learning and exploring deep learning for personal curiosity by doing online courses, personal projects and work.
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Robond Rover Project
Project repository for the Unity rover search and sample return project.
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Hsbm topicmodel
Using stochastic block models for topic modeling
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook
Drl Portfolio Management
CSCI 599 deep learning and its applications final project
Stars: ✭ 121 (+0%)
Mutual labels:  jupyter-notebook

United States 2016 Election results (President, Senate, House, Governor), plus Primaries, Ballot Measures and Exit Polls

These are csvs of results scraped from Politico and (for exit polls) CNN websites. If anyone uses this, I'd love to hear about it! I wrote a brief blog post describing my motivation

For more about the ethics of scraping, see this Quora post. I used Selenium, which automated the Google Chrome browser, so I put no more load on their servers than a normal visitor. The only thing the program did faster than a human is parse the results.

I should address something here: you never *need* Selenium to webscrape, I used it because I 
wanted the results quickly, and there was not much actual data, so the rate limiting step 
was my programmer time. In addition, the code is kind of, well, haphazardly written for the
same reasons. At the same time I did my best to include things like log files to make sure
I didn't go off the rails and to ensure that the results were as accurate as I was able to
produce, which I think is the main point of the exercise.

Note Selenium was handy because, in the case of Politico, the Outer HTML only loaded completely when the browser reached the bottom of the page, so a scraper that could "tell" the website it was really a browser and had scrolled all the way down was necessary. (Again, I could have intercepted the JSON requests, but low-hanging-fruit, Selenium is so easy and I have the API memorized. I'm making such a fuss about this because there's a whole class of beginning web scrapers out there who use Selenium as a webscraping tool when they don't need to, it's a testing tool.)

I make no warranty as to the accuracy of the data, either of the source data itself or of my scraping (although I did my best and tried to validate the completeness of my results). Please contact me if you find any issues not listed here.

I scraped most of the info from Politico, although the exit polls were from CNN. They include:

  • Both General Election and Primaries
  • Presidential, Senate, House and Governor races
  • Ballot Measures
  • Exit Polls
  • By State and by County, where possible.

Further downballot races like state legislatures were not listed.

DATA DICTIONARY:
  • ballot_measure_type: There are 'key ballot measures' and 'state ballot measures'; I am unclear as to the difference between them. Those with the word 'Overall' in this field are key ballot measures.
  • choice: The vote choice for a ballot measure, usually 'yes' or 'no'.
  • delegates: The number of delegates awarded to the winner of a primary. It is null for candidates who did not win the primary. (It would have made sense to replace these nulls with zeroes, but I didn't think of it in time.)
  • fips: The FIPS county code is a five-digit Federal Information Processing Standard (FIPS) code (FIPS 6-4) which uniquely identifies counties and county equivalents in the United States, certain U.S. possessions, and certain freely associated states.
  • geo_name: State name, District name or County name, as appropriate
  • individual_party: For general elections or open primaries, the party of the candidate. For closed primaries, this is blank because all the candidates belong to the same party, which is shown in the party field.
  • is_incumbent: Whether the candidate is an incumbent. The presidential race had no incumbents.
  • is_winner: Whether that candidate won that individual race; in some primary caucuses, where delegates are given out to multiple candidates, even if one candidate receives the most, there is no winner declared per se, so the value of this field is null instead of True or False.
  • name: Name of candidate
  • party: For races in which only one party participated, e.g. closed primaries, the name of that party; otherwise blank
  • rank: 1, 2, 3... signifying a first, second, third... place finish in that individual race
  • reporting_pct: percentage of votes cast that have been reported
  • state: full name of state, or District of Columbia
  • summary: text summary of the ballot measure
  • title: official, short title of the ballot measure
  • vote_pct: percentage of the vote cast in that race that went to that candidate
  • votes: number of votes that went to that candidate

The following are known issues with the data as present on Politico (either the data just does not exist/is not reported, or it exists and for whatever reason, Politico does not have it):

  • Alaska has no Presidential General Election listed by County; perhaps Politico was confused by the fact that they call their counties "Boroughs", although they had no problem with Louisiana's "Parishes"
  • Between the primaries and the election, Shannon County (South Dakota) changed its name to Oglala Lakota County
  • Missing Kansas Presidential Primary by county
  • Missing North Dakota Presidential Primary by county
  • Colorado, Wyoming and Maine Republican primary results are not reported
  • No page exists for Louisiana Senate or House Primary
  • Missing Illinois 5th district Republican House Primary
  • Politico has a "missing something" box in Illinois 12th district House Primary, but both Democratic and Republican primary results are present.
  • Politico's page for Virginia House Primary lists individual district results for only the 2nd, 4th, 5th and 6th districts.
  • Minnesota's Presidential Primary is reported by congressional district, while its Presidential General Election is reported by county
  • Washington's open primaries for senate and house means that by county or district, the primaries are not divided between political parties

If merging this data with county demographic data, be aware of the following:

  • The aforementioned Shannon County change to Oglala Lakota County in South Dakota; fips has changed from 46113 to 46102
  • In 2015, the county of Bedford City, VA, fips 51515, was amalgamated into Bedford County, VA, fips 51019. There is no more fips 51515, while fips 51019 now has greater area and population (and different demographics)
  • Kalawao County, Hawaii, fips 15005, has a population of 90 so they count in a neighboring county's totals (I could not determine which one)
  • As mentioned above, for some reason Alaska's Presidential General Election was not listed per county, and the Presidential Primaries was listed per Congressional electoral district instead of per county. So Alaska has no county data in this dataset. Blame Politico.

I have merged this data, given the caveats above, with Deleetdk's USA.county.data repo; the results are in /merged_with_demog. Note that the demog info also has some political history info, in columns (wide), while my election data is tall/normalized; if you want to compare between elections, that will take some wrangling.

Note that if there is only one candidate with is_winner == True and votes == NaN, they ran uncontested.

A word about the individual_party and party fields. party is filled out when the entire subtable is for one party, e.g. a primary. individual_party is filled out when an individual line showing a candidate lists a political party.

Many of Politico's third-part names were wrong, so the only parties named are "Republican", "Democratic" and "Independent/Other".

Note that the number of votes is reported as a float instead of an integer due to a particularity of the pandas/pydata ecosystem: since they contain NaNs (the uncontested winners mentioned in the previous note), and NaNs are dtype float, the column cannot be integer.

For the curious, the Jupyter notebooks I used to scrape the data are in the notebooks folder. They're not beautiful, but they work. The politico one, in particular, has lots of checks for consistency, both internally and with the demographic data mentioned above.

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].