News

Wilson on MSNBC with Stephanie Ruhle

Oklahoma City Rotary Speech

[imge_pdf_viewer url=http://wpaintel.com/wp-content/uploads/2018/12/OKC-Rotary_VF_181204_2.pdf]

Progressives would ‘lose their minds’ if Clinton ran for president again, says GOP pollster

This article originally appeared on The Hill:


Republican pollster Conor Maguire said in an interview that aired Tuesday on Hill.TV’s “What America’s Thinking” that the progressive wing of the Democratic Party would “lose their minds” if Hillary Clinton launched a third presidential bid. 

“I think the new wing of the party would absolutely lose their minds if Hillary came back and tried to redirect it back towards her,” Maguire, a senior client strategist at WPA Intelligence, told Hill.TV’s Joe Concha on “What America’s Thinking.”

“They have fought so hard to really readjust the party and get themselves back on track, and Hillary would just be an absolute step in the wrong direction for them,” he continued.

Former Clinton adviser Mark Penn and Andrew Stein, a former Democratic Manhattan borough president and president of the New York City Council, both predicted in an op-ed published earlier this week that Clinton would launch a presidential run in 2020.

“True to her name, Mrs. Clinton will fight this out until the last dog dies,” they wrote. “She won’t let a little thing like two stunning defeats stand in the way of her claim to the White House.”

Various progressive candidates, including Rep.-elect Alexandria Ocasio-Cortez (D-N.Y.), have aligned themselves closely with Clinton’s progressive 2016 primary opponent, Sen. Bernie Sanders (I-Vt.). Sanders has been floated as a potential 2020 Democratic contender.

— Julia Manchester

Modeling Black Swan Turnout: Just how rare an event were the 2018 mid-terms?

On election night, something amazing happened.

Outside of a few specialists who study such things, America probably hasn’t come to fully understand how remarkable, and rare, this year’s mid-term elections really were.

Based on data compiled by the brilliant Michael McDonald at the University of Florida, turnout among voting eligible Americans (that is citizens aged 18+, excluding felons who have not had their voting rights restored) is AT LEAST48.5% with some additional remaining ballots unknown (due to counties not releasing how many uncounted ballots they have — and we’re looking at you Broward County, Florida — as well as states where mail ballots merely need to be post marked by Election Day).

The chart below shows turnout rate for every mid-term since 1914. To put this into historical perspective, we have not seen turnout levels for a midterm this high in more than half a century.

We’re now likely to exceed the 1966 mark. That would clearly be a modern record for mid-term turnout. To find a higher mark, you have to go all the way back to 1914, where turnout was 50.4%. We may not get there, but even if we don’t it means we’ve just had a mid-term election with turnout at levels not reached since the early part of the last century. Quite literally a “100-year flood.”

What’s more, some places exceeded their total turnout from the last Presidential race.

Travis County, Texas, for example, which is by far the most liberal major county in the state and includes the University of Texas, cast at least 25,000 more votes this cycle than it did in 2016. Once all of the votes are tallied, it won’t be the only example of turnout in a mid-term exceeding the previous Presidential race. And that’s almost unheard of.

Why was turnout so high?

We’ll probably need to leave it to years of political science research to fully explain why 2018 was such an outlier compared to not just recent, but also distant, history. But, a few explanations are clear from the start.

This election became a referendum on a first term president in a way few mid-terms have in the past. The opposition party almost always has enthusiasm and turns out a lot of votes, but a massive increase in political tribalism magnified that effect.

On the other hand, the President’s party usually sees its turnout ebb, but this year we saw Republican turnout grow and grow, keeping pace with Democrats’ for the most part. President Trump’s ceaseless campaigning was a big part of that-credit where credit is due.

Both 1966 and 2018 also happened against a background of social unrest and upheaval coupled with strong economies. In 1966, the Vietnam War, Civil Rights, and the social pressures that would explode in the late 60s and early 70s were all at the forefront while the economy was humming along with unemployment down to 4%.

2018, too, came with a roaring economy — 3.7% unemployment on October 1, but with an upsurge in protests, demonstrations, and political violence on both sides.

Technology also likely played a role in the surge in turnout this year as both sides increasingly adopted new modes of political communication, incorporating peer-to-peer texting and internalizing the lessons of political science experiments from the last several years with tools like social pressure for turnout efforts.

As pollsters and political data analysts, the real question we have is what comes next?

Accounting for a Black Swan scenario is always a challenge, in part, because a model that would predict a 1-in-100 outcome is going to be wrong most of the time. But, as we look ahead to the 2020 Presidential election and the 2022 mid-terms, we have to ask ourselves what 2018 and history tell us as a starting point to understanding how coming races may evolve.

Even though we may be headed for a 1914 level of turnout, it’s probably a bad comparison to use. In 1914, turnout had dropped in four of the previous five mid-terms. So, 1914 stands as a record only because turnout fell off of a cliff in 1918 and never really reached late-19th/early 20th Century levels again.

The picture in 1966 might prove more instructive. The graph above shows that mid-term turnout moved up in the 60s, spiked in 1966, then held on for one more midterm before dropping precipitously in the 1970s and not really recovering until this year.

The picture in presidential years is similar — turnout was fairly normal for the period in 1964 and in 1968 (both were between 62% and 63% of the eligible population) before dropping into the 50s in 1972 and not really recovering until 2004. From 2004 to 2016, presidential year turnout has remained fairly stable in the low 60s, with a slight uptick in 2004 and a slight drop in 2012.

The typical mode of modeling turnout is to start from vote history and demography and then move to the instrument for changes in interest, intent to turnout, and other factors. These models perform really well when turnout holds within a “normal” range for its historical period — in other words they’ve been quite good for much of the 21st century, for as long as data and technology have allowed us to build them.

But, they don’t really account for the kind of epochal changes that defined a 1914, a 1966, to some extent a 2004, or now a 2018.

So, for the modeler, the question is how we can understand the broader social and historical context to try to predict whether 2020 and 2022 will continue the surge in turnout? Will they be like the 2004 election which was the first in a series of presidential elections with elevated turnout (though the six-point increase from 2000 to 2004 pales in comparison with the size of the shift that happened this year), or will they be like 1966 — a singular surge that quickly digressed to trending downward.

One approach we will explore is if we can build a fundamentals model that controls for the economy, political polarization, measures of social unrest, and other variables that would fit the macro-shifts over time and then include this macro effect in our individual voter estimates.

What history is telling us is that we’ve just experienced the rarest of rare elections. Off the edge of the map is a scary place to be, but it’s also an exciting opportunity to learn something new about how elections unfold.

CNBC: Massive early voting numbers point to record turnout for midterms

Massive early voting numbers point to record turnout for midterms

  • More than 35 million votes have already been cast in early voting nationwide, NBC News’ Data Analytics Lab finds using voter file data from TargetSmart.
  • That figure easily eclipses the total in 2014, when fewer than 20 million early votes were counted on the day before the midterms.
  • Politicians of both major parties frame the midterms as a decisive inflection point on the direction of the country under President Donald Trump.

Early voters are poised to make history.

With just one day to go until hundreds of politicians face off in the midterm elections, more than 35 million votes have already been cast in early voting nationwide.

That figure, found Monday by NBC News’ Data Analytics Lab using voter file data from TargetSmart, easily eclipses the total in 2014, when fewer than 20 million early votes were counted on the day before the midterms.

“We’re well beyond historic midterms now,” said Bryon Allen, chief research officer at political consulting firm WPA Intelligence.

But despite recent surveys indicating that a surge of young voter participation will carry the day for Democrats, TargetSmart’s data from states with early voting suggest that the tallies so far have been virtually split between the two major parties. As of Monday, 42 percent of early voters are Republicans, compared with 41 percent of Democrats and 17 percent with another party affiliation, NBC found.

 The 2018 early vote totals still don’t quite compare with those on the eve of the 2016 presidential election, when more than 46 million early votes were counted. But the still-palpable enthusiasm to vote comes as politicians of both major parties frame the midterms as a decisive inflection point on the direction of the country under President Donald Trump.

“He was a primary factor,” said John Della Volpe, director of polling at Harvard Kennedy School’s Institute of Politics. “One of the legacies of the first term of the Trump presidency is that he has galvanized and inspired millions of young Americans to participate in politics and civic life in ways few people have considered before.”

In a recent survey conducted by the institute, 40 percent of 18- to 29-year-old Americans said they will definitely vote in the midterms. Asked about the slim Republican edge in TargetSmart’s early voting totals, Della Volpe noted that “older Americans tend to vote early and absentee, and they’re more likely to support the president.”

The data also show women outvoting men in key states’ early voting totals, a pattern consistent across the past two biennial election cycles in 2016 and 2014.

Missouri Senate Ballot

[imge_pdf_viewer url=https://wpaintel.com/wp-content/uploads/2018/08/WPAi_MOSEN_Ballot_VF_180822.pdf]

More Bad Polling in Texas

Everything is bigger in Texas: from the critters that roam to the size of the barbecue Texans eat. And we can of course add egregious polling errors to the list.

Later this week, Wednesday to be exact, the Texas Lyceum will release polling on the big four races in Texas (Senate, Governor, Lieutenant Governor, and Attorney General).

Let’s take a trip down memory lane to see just how bad this poll has been. Dating back to 2008 the Texas Lyceum has generously given Democrats a massive house effect boost of seven (7!!!) points.

Let’s dig a bit further as to why they give Democrats such a big boost.

Look no further than their samples as they interview 1,000 adults. But, of course, all adults don’t vote. Likely voters do. But Lyceum interviewing adults leads to independents ranging from 12% to 46% (that’s not a typo — yes, forty-six percent). And Democrats varying from twenty-eight percent (28%) to forty-four (44%) percent. Finally, Republicans have ranged from twenty-three (23%) to forty-four (44%) percent. Pretty radical variance no matter how you estimate party ID in the Lone Star State.

Why stress this? Historically, Lyceum toplines offer very little information and most of their data is buried deep within the crosstabs, which they historically don’t release until after they drop the toplines. While they report the party mix of their adult sample, they don’t consistently report the breakdown of the people who actually responded to the ballot until sometimes days later.

Want another Texas sized error? The Texas Lyceum drastically overestimates the share of Hispanic voters in the Lone Star state. This of course has the impact of decreasing the white vote share under 50%.

Even other polls criticized for how they weight each party know that more than half of voters in Texas are white. Even the Democrat biased UT/Texas Tribune poll from June had the white vote share at 58%.

Given recent public polling and the Lyceum house bias, I expect they show the following: Governor Abbott around +12, LG Patrick and AG Paxton both in a dead heat and Senator Cruz around+3. We’ll see if I’m correct, but if I am, as is likey, I could have saved them thousands!

For anyone interested in reviewing Lyceum past surveys, here are convenient links:

April 3–9, 2017
Results
Cross tabs

September 1–11, 2016
Results
Cross tabs

September 8–21, 2015
Results
Cross tabs

September 11–25, 2014
Results
Cross tabs

September 6–20, 2013
Results
Cross tabs

September 10–26, 2012
Results
Cross tabs

May 24–31, 2011
Results
Cross tabs

September 20–30, 2010
Results
Cross tabs

June 5–12, 2009
Results
Cross tabs

June 12–20, 2008
Results
Cross tabs

April 26-May 7, 2007
Results
Multiple cross tabs for 2007:
Number 1
Number 2
Number 3
Number 4
Number 5
Number 6
Number 7

The Pundit’s Blue Wave in Texas

March 6, as the polls opened across Texas for the Primary Election Day, pundits across the country were breathlessly pointing out correlations between early Democrat voting numbers and the potential of a ‘Blue Wave’ election this coming November.

The self-proclaimed ‘Big Data’ journalists spent a good part of the day, leading up to the polls closing posting heavily caveated tweets and articles alluding to the Democrat early vote advantage across the 15 largest counties in Texas serving as a potential bell weather for Democrat upsets in November. While unsubstantiated observations like this may draw attention to the writer, a closer look at historical data and last night’s results show clear flaws in this overused false narrative.

Our team at WPA Intelligence analyzed the data (as we do), and they show that zero correlation between Democrat primary voting and Democrat performance in the November general elections. Even if Democrats would have had a primary advantage in the 2018 primary (narrator voice: “they did not”), there is no evidence that that impacts the November election.

2008, The Democrat’s Year…

In 2008, a year in which Democrats took control of Congress and the White House, they posted a significant early vote advantage and a significant primary turnout advantage. Despite this fact, they lost both the Senate and Presidential ballot by almost twelve-point margin in Texas.

2002, a Midterm Example

In 2002, Democrats posted a significant primary turnout and early voting advantage. Again, despite the increased energy and enthusiasm, Democrats were outperformed by Republicans at the top of the ballot in both the Gubernatorial and Senate general elections.

2018, Reality

Despite the media’s constant proclamations of record early voting, “huge crowds” and “unprecidented Democrat enthusiasm” signaling party strength, the final election results show that Democrats actually posted a significant deficit in early voting and a deficit that widened still after all of votes were cast on Election Day.

While 2018 Democratic primary turnout represents an improvement from 2014 and 2010 mid-term elections, when compared to the two previous examples it is wishful thinking to argue that this result will magically translate to Democrat victories in November. Democrats have lost major statewide elections in Texas by an average of 15% over the past four cycles. It is not realistic to believe that the March 6th results signal a change in that pattern.

While Democrats have steadily improved in primary turnout in recent years, there is no indication that there will be a ‘Blue Wave’ in Texas this November.

WPAi Wins Campaign and Elections “Reed Award” for Best Use of Data Analytics/Machine Learning in Field Program

Utilizing Analytics for an Optimized Ground Game

The Greg Abbott for Governor re-election campaign entered the 2018 cycle with an aggressive focus on both vote goals and a clear aim to successfully generate Republican-leaning turnout targets at the individual level to support Republicans down the ballot and across the entirety of Texas.

The unique structure of Texas politics, where non-Presidential year turnout is extremely low (averaging 35% over the last three mid-terms) presents an opportunity to identify and turn out a large number of Republican-leaning voters who would typically not participate in a mid-term.

WPAi and our team, led by Senior Data Scientist Dr. Timothy Lee, began our effort to assist the Abbott campaign by building a predictive model of likelihood to vote in the mid-terms and likelihood to support Abbott. This yielded approximately 4.5 million individual targets for turnout efforts.

WPAi then helped the campaign plan the location of 180 field operatives who will be responsible for mobilizing volunteers to contact these targets by using an vertex-covering optimization algorithm.

This algorithm made it possible to effectively build field turfs for all 180 operatives across the massive and diverse expanse of Texas that were both:

1) Geographically contiguous and compact, and
2) Were scaled in such a way that a field operative who the plan called for to be hired 18 months from Election Day would have almost exactly three times the targets as one planned to be hired six months from the election, so that all of the targets could have the highest likelihood of contact.

The first wave of field operatives are already in the field based on the optimized turfs with a second wave planned for early 2018. To date, the campaign has already contacted 232,836 voters at the door.

What was the impact of data and analytics on the campaign?

WPAi’s predictive models and optimized field operative allocation is providing the geographic blueprint for one of the most powerful field efforts ever attempted by a state-level race.

To date, the first wave of field operatives and their volunteers have already contacted 232,836 voters at their doors.

What was the impact of data and analytics on the campaign?

WPAi’s predictive models and optimized field operative allocation is providing the geographic blueprint for one of the most powerful field efforts ever attempted by a state-level race.  To date, the first wave of field operatives and their volunteers have already contacted 232,836 voters at their doors.

The pre-built optimized turf structure also allowed the campaign to adapt seamlessly after Hurricane Harvey inundated much of East Texas. Operatives were quickly re-assigned or hired to cover other turfs in unaffected areas of the state while Houston and the Gulf Coast recovered.

Pictured: The WPAi team (left to right) Conor Maguire – Senior Client Strategist, Ashlee Rich Stephenson – Chief Strategy Officer, and Chris Wilson – CEO.

About Dr. Timothy Lee, the mastermind behind this Optimized ground game: Tim is an expert in mathematical optimization and his current interest lies in applications of convex optimization to machine learning and statistical modeling. When Tim is not in the office, you will find him at local tea shop studying mathematics and reading Milton Friedman. He was once known as “the sniper” for his remarkable jump shot in Basketball. Tim holds a Ph.D. in applied mathematics from Rensselaer Polytechnic Institute, specializing in operations research.

Chris Wilson Appears on FNC’s ‘Happening Now’ To Discuss Trump Poll Numbers