Why Did Clinton Lose? Why Did Trump Win? Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 Why Did Clinton Lose? Why Did Trump Win? Milo Schield Augsburg College Editor: www.StatLit.org US Rep: International Statistical Literacy Project TC Metro Critical Thinking Club www.StatLit.org/pdf/2016-Schield-CTC2-Slides.pdf www.statlit.org/pdf/2016-Schield-CTC2-Exit-Poll-Summary.pdf 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 1 1

We Want Explanations: Wrong Color Pantsuits? Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 2 We Want Explanations: Wrong Color Pantsuits? . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 2 2

Plausible Explanations Must be Supported by Data Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 3 Plausible Explanations Must be Supported by Data . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 3 3

FBI Comey Letters . 2nd 1st 4 Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 4 FBI Comey Letters . 2nd 1st 2014-Schield-ICOTS-6up 2008SchieldNNN6up.pdf 4 4

What if only ____ voted? Women vs. Men Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 5 What if only ____ voted? Women vs. Men Source: www.motherjones.com/politics/2016/10/maps-presidential-election-race-gender-age 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 5 5

What If Only ___ Voted? People of Color vs. White Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 6 What If Only ___ Voted? People of Color vs. White . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 6 6

What If Only ____ Voted? Whites: Women vs Men Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 7 What If Only ____ Voted? Whites: Women vs Men . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 7 7

What if only ____ Voted? Whites: College vs. No College Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 8 What if only ____ Voted? Whites: College vs. No College . Data Source: Nate Sliver (538) in mid October, 2016. 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 8 8

Exit Polls: Poll actual voters Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 9 Exit Polls: Poll actual voters . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 9 9

Exit Polls Voters Asked on Exit Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 10 Exit Polls Voters Asked on Exit PLUS: Best way to contact actual voters. Connect candidate choice with multiple groups. No undecided after voting Get smaller subgroups. MINUS: Not necessarily representative. Sampling bias? Some chose “None of the options” or “No answer” Subject bias is still relevant. 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 10 10

Exit Poll Results: Questions with Two-Answers Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 11 Exit Poll Results: Questions with Two-Answers What groups should each candidate focus on? Pro-Clinton Groups (Most are for Clinton): Sort by Group’s share of population Sort by Clinton’s share of group vote www.statlit.org/pdf/2016-Schield-CTC2-Exit-Poll-Summary.pdf Source of Exit Poll Data: www.cnn.com/election/results/exit-polls www.nytimes.com/interactive/2016/11/08/us/elections/exit-poll-analysis.html 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 11 11

Pro-Hillary Exit-Poll Sheet Left Side Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 12 Pro-Hillary Exit-Poll Sheet Left Side 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 12 12

Pro-Hillary Exit-Poll Sheet Right Side Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 13 Pro-Hillary Exit-Poll Sheet Right Side 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 13 13

Multi-Group Choice-Based Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 14 Multi-Group Choice-Based 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 14 14

Multi-Group Opinion-Based Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 15 Multi-Group Opinion-Based 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 15 15

Multi-Group Choice or Opinion-Based Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 16 Multi-Group Choice or Opinion-Based 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 16 16

Multi-Group: Fixed Conditions Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 17 Multi-Group: Fixed Conditions 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 17 17

Critical Thinking Review What Other Data Would Help? Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 18 Critical Thinking Review What Other Data Would Help? We have lots of data comparing Clinton & Trump. Two-answer data (See Handout) Multiple-answer data (3-7 choices) Is this data sufficient/adequate for our decision? If not, what other data would be helpful? We need time-data: Hillary compared to Obama. It takes a change to explain a change. 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 18 18

Time-Compare by Gender: Increase among Men Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 19 Time-Compare by Gender: Increase among Men . 20 point gender gap 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 19 19

Time-Compare by Race: Trump bested Romney in All Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 20 Time-Compare by Race: Trump bested Romney in All . 50 point race gap 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 20 20

Time-Compare by Income: Trump better in Low-Income Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 21 Time-Compare by Income: Trump better in Low-Income . 15 point income gap Confounded by race 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 21 21

Time-Compare Among Whites by Education Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 22 Time-Compare Among Whites by Education . 30 point education gap: After controlling for race 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 22 22

Biggest Problem: Confounding Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 23 Biggest Problem: Confounding 50 point gap by Race (70% White vs. 30% Non-White) 20 point gap by Gender (48% Men vs. 52% Women) 15 point gap by Income (Low vs High: Low is less than $50K) 5 point gap by Education (50% no College vs. 50% College) Statistical Literacy: Take into account most important factors first. Otherwise the results will be “confounded”. Simplest way: Selection! 30 point gap by Education among Whites (no College vs. College) 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 23 23

Women: White, High School or Less Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 24 Women: White, High School or Less . 14 pt gap Control for Race and Education by Selection Source: PEW Research Center 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 24 24

Men: White, High School or Less Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 25 Men: White, High School or Less . 38 pt gap Control for Race and Education by Selection www.npr.org/2016/11/08/501069216/5-exit-poll-numbers-to-watch-on-election-night 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 25 25

Psychological Explanation: Trump is a Populist Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 26 Psychological Explanation: Trump is a Populist . 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 26 26

Psychological Differences: Trump=Populist; Clinton=Not Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 27 Psychological Differences: Trump=Populist; Clinton=Not Source. Oliver and Rahn (2016). Rise of Trumpenvolk: Populism in the 2016 Election. Annals AAPSS 667, Sept 2016. URL: http://ann.sagepub.com/content/667/1/189.full.pdf Trump Clinton Sanders 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 27 27

Critical Thinking Summary Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 28 Critical Thinking Summary What kind of data do we need? “It takes a change to explain a change” Time-based data is essential If data doesn’t include the motivating reason or condition, that data will not be helpful. Let’s look at some psychological “data”. 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 28 28

“Trumpland” Excerpt by Michael Moore Schield ICOTS Analyzing Numbers in the News 15 May 2008 2014 29 “Trumpland” Excerpt by Michael Moore https://www.youtube.com/watch?v=YKeYbEOSqYc Uploaded Oct 24 2.3 million views as of 12/03 Why Trump will win 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 29 29

Michael Moore “Trump’s election is going to be the biggest Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 30 Michael Moore “On Nov 8, the dispossessed will walk into the voting booth and … vote … for the man that has threatened to upend and overturn the very system that has ruined their lives: Donald J. Trump” “Trump’s election is going to be the biggest ‘fuck you’ ever recorded in human history” “And it will feel good !!!” 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 30 30

Why did Moore Get it Right? Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 31 Why did Moore Get it Right? He had a personal relationship with the respondents. Moore: “I know a lot of people in Michigan that are planning to vote for Trump and they don't necessarily agree with him. They're not racist or redneck, they're actually pretty decent people. After talking to a number of them, I wanted to write this. “We went to union halls and guys that I grew up with — people who normally vote Democrat who are thinking of voting for Trump. That is a huge chunk of the population, especially where I'm from.” 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 31 31

Critical Thinking Summary Analyzing Numbers in the News Schield ICOTS 15 May 2008 2014 32 Critical Thinking Summary CTC1: Election poll forecasts -- closer to fortune telling CTC2: Why Clinton Lost. Looking at the wrong data can be debilitating! It takes a change to explain a change ‘Movie’ data is better than ‘snapshot’ data Take into account (control for) the biggest factor first Simplest way to control for something: Selection 2008SchieldNNN6up.pdf 2014-Schield-ICOTS-6up 32 32