Questions

Week 3

-Media diet assignment due on Monday July 6 -Discussion Week 3: first post due on Wednesday July 8, second post due on Sunday July 12 -Three lectures and two readings: https://fivethirtyeight.com/features/the-real-story-of-2016/ https://www.politico.com/news/magazine/2020/04/12/coronavirus- 2020-campaign-178487 -Coming for week 4:

• Proposal for final assignment due on July 15 (proposal and final assignment description are uploaded on Canvas under assignments and will be discussed more detailed in the lectures for week 4)

Polling and Forecasting

What’s the problem with polling?

The Infamous 2016 forecast maps

Are polls becoming less accurate? •Polls have not become worse

•On average, they are as accurate as ever

•Polling mistakes can happen

•Mass media reporting on polling is full of inaccuracies

5 Steps to conduct apoll 1. Define the universe

• Entire population the poll will measure (women, men, general population, specific ethnic/racial groups, etc.)

2. Construct the Sample 3. Develop valid and unbiased questions

• Must be very specific and non-leading 4. Control how the poll isconducted

• Phone, e-mail, face-to-face (careful that the interviewer doesn’t influencethe answers)

5. Analyze and report the findings

Basics of samples •Random Sampling: principle that everyone should have an equal probability of being selected for the sample

• Representative Sample: Draws from all groups •Sampling Error: The level of confidence in the findings of a public opinion poll, +/-3%

•Example: • Poll that says that Trump is at 43% and Clinton at 42%, with plus/minus margin of

error • What’s the maximum/minimum percentage that each candidate could have?

Margin of error example

When you’re reading a single poll, remember these 4 points:

•1) There is always a MARGIN OF ERROR in the sample (don’t just read the predicted share of the vote, read the margin of error)

•2) Always look at the type of poll • Registered voters/Likely voters • Online/Phone • Sample size/representativeness • Republican leaning/Democratic leading

•3) Polls give you a snapshot of the state of the race at that particular moment •4) Be careful and look at the number of undecided and whether they are counted in the final poll numbers • Are they excluded from the final numbers? • Are they allocated 50/50 to each candidate?

Why polling can be useful ◦ Help candidates and politicians figure out public preferences ◦ Help candidates decide where to spend resources during a campaign ◦ Inform the public about issues ◦ Exit Polls- used by the media to predict election day winners so we don’t have

to wait! ◦ Polls reflect the policy agenda—problems the people inside and outside of

government believe must be addressed

Negative side of using too muchpolling •Focus on politics as a game and horserace rather on substantive issue •Write off some candidates when they have achance

•If there is a polling mistake, campaigns can get the wrong message

•Could demobilize or mobilize some voters

•Politicians rely too much on polling in decisions, rather on what’s best for the country/state/local community

How “forecasting” works •Aggregate of polls

• All credible polls in a particular race

•Consider partisan leaning of polls • Some polls are done by firms closed to the Republican or Democratic party

•Consider quality of polling

•Consider the “fundamentals” in the race • Economic data, approval for the each party of the candidates, popularity of candidates

How forecasting works •Use all these data to come up with aMODEL

• What is the PROBABILITY that a candidate could win the race • Probability are difficult to grasp: a 30% is not that low as probability •Coin with two sides (one Trump/one Clinton): 30% probability means that 3 times out of 10 the coin will flip to Trump

How mass media use polling •Report on who’s up and down •Forecasting the results of the election •Narratives about who are the losers and the winners (before the election) •Discuss political strategies that can address polling issues for candidates •Don’t talk about uncertainty and margin of error very well

What happened in 2016? 4 factsabout polling

1. Most national polls indicated that Clinton was aheadNATIONALLY 2. State level polling in swing states was sparse and contradictory and there

were not many in the lastweek ◦ Post election research: polls were underestimating the turn out of white

working class people in the Midwest

3. A lot of undecided voters decided who to vote for in the last week ◦ In some swing states there were 20% of undecided voters ◦ The majority moved towards trumpduring the last week

4. There was a last minute event ◦ Comey letter that stated that the FBI investigation in Clinton’s email was reopened

Did the Comey letter change the outcomeof the election and affect polling?

•Comey letter came out on October 28 • Reopening the investigation in the Clinton’s emails

•Polling didn’t really catch up with the changes in candidate support •National and state level polling showed a decreased in support

for Clinton

•In many swing states undecided went for Trump • 70% for Trump, 30% for Clinton

•Mass media focused extensively

on the letter in the last week

Media narrative versus data •Media LOVE certainty

• Polling and forecasting have A LOT of uncertainty

•Media don’t know how to talk about probability • 30%/40% probability is not extremely low

•They used a narrative and stick with it • Don’t look at signs that could disprove their narrative • Don’t look at outlier polls that tell a different story

•Bias/Standardization • Lack of understanding of areas of the country/specific populations that decided the election • Mass media do not understand rural America very well

What can mass media do? •Explain how polling works

•Be more accurate

•Have reporters that understand

different parts of the country

•Be more transparent

•Analyze polling critically

•New creative ideas • Partner with polling organizations