2016 Presidential Election Polls Analysis
11-14-2016 20:32 PM - last edited 01-09-2017 11:15 AM
11-14-2016 08:32 PM - last edited on 01-09-2017 11:15 AM by mackenzie_lyng
If the embedded report is not showing above, try this link: https://app.powerbi.com/view?r=eyJrIjoiZmMyOTljZTYtMjIxOS00YjgxLWI2NzEtZjFkMDEzZWM4ZDYyIiwidCI6IjRhM...
Perhaps inevitable, there's just too much interesting data out there around the U.S. 2016 Presidential Election. One of the key narratives coming out of this year's U.S. Presidential Election is that "the polls were wrong". As with any generally accepted narrative, let's see what the data says. Were all of the polls wrong? A majority wrong? How wrong were the polls? Were different types of polls more accurate than others? Fear not, this data story answers those questions and many more! Find out the poll "winners" and "losers". Filter slice and dice by state or general election polls or by individual polling organizations. Find out if all polls are created equal. Four years from now, you'll know which polls to trust and which polls you should ignore, thanks the power of Power BI!
The key metric used in the report is "Error". Error is simply the absolute value of the difference between what the poll predicted (predicted margin) and what actually happened (margin). For example, if the poll predicted Clinton as the winner by 3 percentage points and Trump won by 1.5 percentage points, then the Error would be 4.5 percentage points (denoted 4.5 %). Please note, % is used as shorthand for percentage points in this report.
Key insights include:
- Individual state polls were 1.5 times worse than general election polls
- Self-identified "tracking" polls were 1.8 times better than polls that were not identified as "tracking" polls
- Polls that predicted Trump as the winner were actually less accurate than polls predicting Clinton as the winner due to many state polls predicting Trump severely underestimated his margin of victory in those states
- While the average % Error was trending down for all polls (both state and general), general election polls average % Error was actually trending up as the election came closer
- Across all polls, Y2 Analytics had the worst average error at 13.99% predicting Utah
- Across all polls, Winthrop had the most accurate poll, predicting Virginia within 0.19%
- Of the national (general election) polls, NBC News/SM was the worst poll with an error of 6.31%
- The IBD/Tipp Tracking poll predicted the general election most closely with an average error of just 0.94%
- Monmouth had the most monumental miss (in a single individual poll) with an error of 18.25% in Indiana.
Page 1: Overview provides a quick snapshot of the polls, including the poll "winner" and "loser" using the new Top N filter. The user can look at all polls, just general election polls individual state polls.
Page 2: Users can drill into state or general election polls as well as each individual polling agency.
Take a look and enjoy!
- Poll data was obtained from RealClearPolitics.com. Election results data was obtained from Wikipedia.org
- For "winner", only popular vote is considered as this is what the polls measured
Did I answer your question? Mark my post as a solution!
Proud to be a Datanaut!