Pollsters redeem themselves after 2016 debacle

Logan Hullinger
York Dispatch

Two years ago, a pollster famously said he'd eat a bug if Donald Trump was elected president.

Days after the 2016 presidential election, Sam Wang, founder of the Princeton Election Consortium, downed crickets dipped in honey live on CNN after Trump proved him and most other pollsters wrong.

Those polls were so flawed that some Democrats were reluctant to believe surveys showing their party taking control of the House in this week's midterm election

As it turns out, polling this election season was mostly on the mark, at least for state and local races.

Midterm results: For example, a wide range of polls showed Democrats Gov. Tom Wolf and Sen. Bob Casey with double-digit victories.

The polling aggregation site RealClearPolitics averaged 11 local polls to calculate a 19.6-point win by Wolf. Political analytics websites such as FiveThirtyEight and The Cook Political Report use such polls as a factor in their election forecasts.

More:Gov. Tom Wolf wins re-election bid against Wagner

The Cook Political Report called the race "likely Democrat," and FiveThirtyEight called it "solid Democrat" with an estimate of a roughly 15-point lead for Wolf.

On Tuesday, Wolf easily beat Republican former state Sen. Scott Wagner to earn another four-year term in office. Wolf received 57.6 percent of the votes compared to Wagner's 40.8 percent, a difference of 16.8 points.

As in the gubernatorial race, RealClearPolitics aggregated 11 polls and predicted a 12.8-point lead for Casey in the Senate race. FiveThirtyEight predicted roughly an 11-point lead.

Casey won the race Tuesday with 55.6 percent of the votes compared to Republican Lou Barletta's 42.8 percent, a 12.8-point margin of victory.

More:Bob Casey beats Trump-backed Barletta in Senate race

U.S. Sen. Bob Casey, D-Pa., gives a victory speech during his election night party in Scranton, Pa., on Tuesday, Nov. 6, 2018. (Jake Danna Stevens/The Times-Tribune via AP)

In addition to those races, three-term Republican incumbent Rep. Scott Perry was estimated to earn a slim victory in the 10th Congressional District, and one-term Republican incumbent Rep. Lloyd Smucker showed a comfortable lead in the 11th District.

In the 10th District, the only polls conducted were from the left-leaning polling firm Public Policy Polling, which conducted two polls. The first showed Perry with a 4-point lead, and the second showed him with just a 1-point lead.

The Cook Political Report called the race a "Republican toss-up," favoring Perry but giving Scott a chance. FiveThirtyEight rated the district as "lean Republican." 

Perry, of Carroll Township, narrowly earned a fourth term Tuesday over Democrat George Scott by pulling in 51.4 percent of the votes compared to the Democrat's 48.6 percent, a 2.8-point difference.

Vice President Mike Pence rallies with candidates Scott Perry, left, and Lloyd Smucker during Pence's stop at Lancaster Airport in Lititz Wednesday, Oct. 24, 2018. Bill Kalina photo

More:Rep. Scott Perry edges out George Scott in 10th District race

On the other hand, polling for the 11th District was less accurate.

The 11th District race was only polled once by PPP, which underestimated the race and gave Smucker a 9-point victory. A more recent poll conducted by Susquehanna Polling and Research for ABC27 News showed Smucker with just a 4-point lead.

However, Smucker easily kept his seat red against Democrat Jess King Tuesday, bringing in 59 percent of votes compared to King's 41 percent, an 18-point victory.

With more realistic foresight, The Cook Political Report didn't even list the race as competitive, and FiveThirtyEight gave the race a "likely Republican" classification and underestimated the results with a 9-point lead for Smucker. 

More:Republican Rep. Lloyd Smucker beats out Democrat Jess King for second term

Polling strategies G. Terry Madonna, director of the Franklin & Marshall College Center for Politics and Public Affairs in Lancaster, said polling is an important way to gauge public opinion if done well.

F&M conducted several polls on the state's gubernatorial and Senate races, all of which proved accurate come Election Day.

The reason for their accuracy, he said, was the polls' methodology.

Madonna said there are four steps to conducting accurate polls. First, a "good, randomly selected" group of people that represents the region's voters well is needed.

Director of the Franklin & Marshall College Poll and Professor of Public Affairs at Franklin & Marshall College Dr. G. Terry Madonna leads a political discussion, 2016: Presidential Politics in the Nation and State, at Yorktowne Hotel in York City, Tuesday, Sept. 13, 2016.

Next, the polling firm needs questions that are "easily understood and not biased," and then it needs to make sure statistical adjustments, such as margin of error, are noted and properly formulated.

Lastly, the entity needs to contact voters and speak to them or give them online surveying options instead of using automated polling methods, which can make finding an adequate pool of survey candidates difficult, Madonna said.

After these steps are followed closely, polling can prove to be accurate, as was the case this year in Tuesday's midterm elections, he concluded.

"The pre-election polls were very accurate," Madonna said. "I was reasonably confident we would be within the margin of error (of the average poll results)."

More:Dem and Republican leaders: Midterm enthusiasm high in York County

What the results showed, he said, was unity within the Democratic Party, which helped the Democrats gain power in the House.

"Democrats were more cohesive," Madonna said. "People were more likely to vote straight party. That wasn't the case with Republicans."

Still, state races in counties such as York went against the grain of larger trends. Every state-level York County Democratic candidate lost their contested races, none of which was polled beforehand. State Rep. Carol Hill-Evans, D-York City, was running unopposed.

Madonna said the more local results came about because the county is predominantly conservative, which "Republicans, specifically at the state level, rely on."

The most recent Department of State statistics show there are 151,270 registered Republicans in the county compared to 103,669 Democrats — a 37.3 percent difference.

— Logan Hullinger can be reached at lhullinger@yorkdispatch.com or via Twitter at @LoganHullYD.