Since we have to wait 138 days to find out the composition of the 115th Congress, polls and takezzzzz will have to suffice. The “Blue Wave” is either “alive and well” or “crashing and burning.” Panic abounds when the Democratic generic House vote margin expands or shrinks. Currently, you can win a dollar by betting 62 cents on the Democrats to win the U.S. House and make a nice profit if you lay some cash down on the Republicans. So, are the Democrats going to win or not? It’s complicated.

In 2014, the Democrats did abominably. They lost 13 seats and fell to a decades low 188 members. Republicans captured 247 districts for a 59 seat advantage. Nationally, in the exit poll, Democrats lost by six points (47% to 53%). Given the historically low ratings of Donald Trump, Democrats should expect to do better than in 2014. However, looking at the current share Democrats receive in the daily Civiqs survey, Democrats are almost exactly where they were when they got croaked.

Democrats get over 90% from their partisans. Almost no Republican crosses over. Democrats do better with Independent women than men. The party has not changed any minds since 2014. For this distribution to produce a majority in the House, Democrats need to drive up turnout with their supporters and hope the Republicans fail to catch up in base mobilization. Since hope isn’t a plan, Democrats need to persuade Independents.

The last time the Democrats won the House, in 2006, they got 53% of the voteand took 233 seats. President Bush had an approval rating of 37%, a few points less than where Trump is now. Democrats are running 22 points behind with Independent men, 15 points behind with Independent women, and five points behind with Republican men. They are at the same level now with Democrats and Republican women. To overcome gerrymandering and build a majority large enough to govern, Democrats have to dramatically improve with Independents.

Of course this is just comparing Democratic vote share, not the margin over the Republicans. However, I have been part of and observed too many campaigns where the Democrat finishes with the same percentage they got in the last poll. Although this wasn’t a problem in my polls in 2016, polling has gotten worse at capturing the views of white rural and non college educated voters. In Kentucky in 2015, Democratic pollsters went through their pre-election polls for Governor (Democrat Jack Conway led by about six points in pre-election polls and lose to Republican Matt Bevin by nine points) and determined that most of the undecided voters broke to the GOP (no link, I heard this in a briefing).

This all occurred when Barack Obama was President. He’s not anymore, so let’s look at the Republicans. The Republicans are far off from their 2014 stomping. They trail by five to six points among both Republican and Democrats. They do worse with Independents — 19 to 20 points off. Not good.

When the Republicans lost in 2006, they kept their own voters, got almost no Democrats, and just picked up two out of every five Independents. Compared to then, Republicans are doing about the same with Republicans and Independent men and are a few points off among Republican and Independent women.

Whichever side can win over Independents will control the House. Republicans currently lead with Independent men:

While Democrats lead with Independent women:

Independents really don’t like either party. Independent men are favorable towards Trump while women are not.

Civiqs does not allow me to refine these results in order to isolate Independents who are undecided in the race for the U.S. House or support another party. My guess is that these 20% are even more unfavorable towards both parties and the President.

The election will ultimately swing on which party these voters hate more on Election Day. Stay tuned!

Like the proverbial broken clock, the New York Times is right today. The 2016 exit polls did lead us to misinterpret the 2016 election. Voters were less educated and older than what the exit polls told us. White non-college voters do make up a greater share of the Democratic electorate than is commonly reported. The article goes into great depth on the methodological problems with exit polls that we should just all accept so we can move on from using exits as reliable guideposts in the future. Unfortunately, this fuller understanding of the Democratic coalition should not lead us into pursuing messaging and issues strategies that will break apart that coalition or dampen the considerable enthusiasm building for the 2018 and 2020 elections.

After presenting the data, Thomas Edsall quotes the academic William Galston about what the Democrats should do as a result:

Galston, writing in the March 16 Wall Street Journal, argues that Democrats need to moderate their stand on immigration in order to win over white noncollege voters.

“No issue has done more than immigration to feed populism, and finding a sustainable compromise would drain much of the bile from today’s politics,” Galston writes. He continues:

Defenders of liberal democracy should acknowledge that controlling borders is a legitimate exercise of sovereignty, and that the appropriate number and type of immigrants is a legitimate subject for debate. Denouncing citizens concerned about immigration as bigots ameliorates neither the substance nor the politics of the problem. There’s nothing illiberal about the view that too many immigrants stress a country’s capacity to absorb them, so that a reduction or even a pause may be in order.

First off, voters really hate Trump’s immigration policies. In a February CNN poll, 36% approved of them while 60% disapproved, identical to his overall approval rating in that poll (35% approve-58% disapprove). A substantial majority (60%) oppose his signature proposal — the “Big Beautiful See-Through Wall.” Only 38% approve of this, less than the percentage of people who voted for him in 2016. On the other hand, 80% support allowing undocumented immigrants who were brought to the U.S. as children to remain in the United States and eventually apply for citizenship.

Secondly, how are Democrats supposed to “moderate” without losing the support of the “white working class” voters who supported Hillary Clinton, not to mention white college graduates and people of color? Edsall cites Galston’s work in June 2016 that showed the major differences between white college graduates and white non-college graduates (pollsters and pundits use “white working class” as short-hand for not graduating from college. Plenty of college graduates, even post-graduates, work as well but “white non-college graduates” is a clunky phrase) on a variety of issues. White non-college graduates hold much more conservative positions than white college graduates. That shouldn’t surprise anyone as they are more conservative and more Republican to begin with.

You know what else divides Americans? Partisan identification. It’s even more predictive of vote than race and educational attainment.

Looking at policy preferences by race and education doesn’t mean much unless it is also analyzed by party. If white non-college Democrats agree more with Republicans, both college and non-college, than white college Democrats, the party should look to “moderate” on core civil and human rights issues. They don’t however. The chart below shows results on a variety of issue questions from the Cooperative Congressional Election Study (CCES)in 2016. The chart is kind of messy but it breaks out five groups:

• White Non College Trump Voters (light red)
• White College Trump Voters (dark red)
• White Non College Hillary Voters (light blue)
• White College Hillary Voters (dark blue)
• Non White Voters (orange)

On every measure, Hillary voters — both college and non-college — agree with each other and not with Trump voters. This covers a wide variety of issues: abortion, criminal justice, environment, government spending, guns, immigration, racial issues, health care, and others.

On some issues, white non-college Hillary voters are slightly more conservative than white college Hillary. On the other hand, white non-college Trump voters are slightly more liberal than white college Trump voters on issues like the minimum wage, increasing health care spending, and on environmental policies. Democrats have more to worry about with non-white voters than they do the white non-college voters who support them.

Edsall also quotes Ruy Teixeira of the Center for American Progress:

There is no way around it — if Democrats hope to be competitive in Ohio and similar states in 2020, they must do the hard thing: find a way to reach hearts and minds among white non-college voters.

Well, they have reached the hearts and minds of some white non-college voters. Also, 54% of the electorate did not vote for Trump. By “moderating” on immigration, particularly by supporting draconian policies like the “Wall” or breaking up families, Democrats risk turning off the non-college white Hillary voters along with depressing turnout with white college Hillary voters and non-whites. And for what gain? The white non college Trump voters are vastly more conservative on issues other than immigration and would not vote for Democrats anyway. Unless you can prove that by changing policy positions, values, and issue emphasis you will gain more votes than you will lose, forget about doing it. Just saying that there are more white non-college voters in the electorate than previously thought does not solve this equation.

Much is made of the Obama-Trump voter but they are vastly more conservative than the Obama-3rd party voter:

This is not to suggest that Democrats abandon the white non college voter. They should just not change policy preferences on core issues involving civil and human rights. The actions by Trump’s Immigration and Customs Enforcement (ICE) agents are abominable.

A vast majority of voters support immigration reform that provides a pathway to citizenship for all illegal immigrants and oppose Trump’s wall. They also believe immigrants strengthen the country through their hard work and talents (65%) while a small minority (29%) believe immigrants burden the country by taking jobs, housing, and health care. Democrats have nothing to gain by turning against their own coalition no matter how many people in Washington and the media say they should. They should just nominate better candidates.

Last week, after being roasted in a town hall with survivors and victims of the mass shooting in Parkland Florida, Senator Marco Rubio took to Twitter to denounce an assault weapons ban as “outside the mainstream.”

Banning all semi-auto weapons may have been popular with the audience at #CNNTownHall, but it is a position well outside the mainstream https://t.co/18rMDwEfGs

— Marco Rubio (@marcorubio) February 22, 2018

For someone who has spent a small fortune on polling in his political life, Senator Rubio certainly doesn’t know how to read one. Just yesterday, Florida Atlantic University released a poll showing that 69% of Floridians want a ban on assault rifles. Many other pollsters have found similar results nationally.

While we can expect support for common sense gun laws would spike after a horrific mass shooting, it’s instructive to go back to a neutral point in recent history to get a calmer read. The Cooperative Congressional Election Study (CCES) does a massive poll during and after federal elections. In 2016, they asked the following:

On the issue of gun regulation, do you support or oppose each of the following proposals? Ban assault rifles

Among the 44,169 registered votes surveyed who voted in the general election, 63% favor banning assault rifles. This support is deep among many groups in the electorate.

Looking at the electorate by race, education, and county type, all voters in large urban counties overwhelmingly support the assault weapons ban. People of color in every type of county — urban, rural, and everywhere in between — support the assault weapons ban. Among white voters, college graduates in all type of county support it. Non college white voters in small towns split and rural areas narrowly oppose it.

White women — both those who have graduated from college and those who have not — support the ban. White non college men in urban areas split while those in more rural areas oppose it. White college men in large and medium population counties support the ban. Those in small town counties split, and those in rural counties are opposed.

In total, white women, people of color, and white college men support a ban. That’s the very definition of “mainstream.” While “rural voters” are the excuse members of Congress and pundits give for why any laws to deal with gun violence can’t pass, rural white male voters represent a small portion of the population and don’t have enough members in Congress to thwart a majority. Don’t blame “rural voters” for inaction. Blame politicians who ignore the will of the vast majority of their electorate.

Even though Trump voter in every region oppose the assault weapons ban, a significant share in urban areas, over 40%, support it. Big city Trump voters like this guy:

As far as the “rural” areas, a third or more of Trump voters in rural and small town counties favor the assault weapons ban. These voters might be out of the “mainstream” of the communities in which they live, but not with voters in the rest of the country. More Hillary voters support the ban than Trump voters oppose it. More than 80% of Hillary Clinton voters in every region favor this ban. Democrats will not “abandon” the party if members vote for a ban, even Democrats outside of cities. Third party voters are largely split — unsurprising given the frothy mix of libertarian Gary Johnson voters and far-left Jill Stein supporters.

Support for the Assault Weapons Ban by State

Looking at individual states, banning assault weapons is a majority position in every state except Alaska, Idaho, Montana, Oklahoma, West Virginia, and Wyoming. We calculate this using the CCES data and a technique called Multi-Level Regression and Poststratification (MRP). This allows us to estimate support at the state and congressional district level despite small sample sizes. This technique is widely used with CCES data. The datasets and R code used can be found here.

Overall, 88 senators represent states where 50% or more voters support an assault weapons ban. 50 senators represent states where 61% or more voters support an assault weapons ban. If senators voted according to the majority of their electorate, an assault weapons ban would pass with ease even with the “rural state bias.”

Assault Weapons Support by Congressional District

If members of the House voted for the majority position on an assault weapons ban in their district, it would pass 399 to 36. Democrats would vote 192 to 2 and Republicans 207 to 34.

Increasing the threshold to 60%, the ban passes 271–164. Democrats would vote 173–21 in favor. Republicans would vote 98–143 in opposition. In 109 Republican districts, support for the assault weapons ban is between 50% and 60%. Who again is out of the mainstream?

The vast majority of districts (196) that voted for Trump support the assault weapons ban. Only 34 oppose it. Voters in only one district held by a Democrat (AZ-01, Tom O’Halleran) and won by Trump opposes the ban. The 23rd district in Texas, represented by Republican Will Hurd, was won by Hillary but voters there oppose the ban.

A substantial number of districts represented by Republicans but voted Hillary strongly support the assault weapons ban. These include:

CA-39 — Ed Royce — 77.4%
FL-26 — Carlos Curbelo — 75.8%
TX-32 — Pete Sessions — 75.1%
TX-07 — John Culberson — 71.5%
NY-24 — John Katko — 70.3%
CA-21 — David Valadao — 70.0%
IL-06 — Peter Roskam — 69.1%
AZ-02 — Martha McSally — 69.1%
CA-49 — Darrell Issa — 68.5%
PA-07 — Pat Meehan — 68.0%
WA-08 — David Reichert — 67.6%
CA-48 — Dana Rohrabacher — 66.9%
CA-45 — Mimi Walters — 66.2%
FL-27 — Ileana Ros-Lehtinen — 66.1%
MN-03 — Erik Paulsen — 66.0%
CA-25 — Steve Knight– 66.0%
NJ-07 — Leonard Lance — 65.3%
CA-10 — Jeff Denham — 63.2%
CO-06 — Mike Coffman — 61.7%
PA-06 — Ryan Costello — 60.7%
KS-03 — Kevin Yoder — 56.7%
VA-10 — Barbara Comstock — 55.7%

Many of these members are retiring or are prime targets for Democrats in 2018. Democrats running in these districts have nothing to fear from supporting an assault weapons ban.

In a normal democracy, where representatives vote the will of their voters, the AR-15 and other weapons would no longer be legal to buy. However, we don’t live in a normal democracy. Just don’t let anyone tell you that your belief that assault weapons should be banned is out of the mainstream. They are misinformed or are lying to you. All this data backs that up.

I was listening to Laurence O’Donnell on the David Axelrod podcast and they talked about race issues in Boston in the 1970s and how Ted Kennedy had real problems in parts of Boston because of his support for busing.

Looking at Kennedy’s Senate results over time shows this. In 1970, three times as many Boston voters left the U.S. Senate race blank (15.8%) compared to voters statewide (5.3%). In 1976, twice as many Boston voters left the U.S. Senate race blank (8.1% Boston-4.0% Statewide). This difference narrowed over time except for a spike in 2000.

In 1970, Kennedy did only 2.5 points better in Boston than he did statewide. In 1976, he did worse in Boston (63.6%) than he did statewide (66.6%). In later elections, Kennedy expanded his support every year and significantly outperformed in Boston than he did in the rest of the state.

Racial issues poisoned Boston in the 1970s and Ted Kennedy was not immune from it.

The US Senate has taken 36 votes so far in 2018. They passed a budget, failed on Dreamers (so far), and confirmed many Trump nominees. The chart below shows the groups of senators who vote most often with their colleagues. Various groups on both sides of the aisle emerge. These are senators whose votes correlate more than 90% of the time.

The main Republican block includes Thom Tillis (NC), Orrin Hatch (UT), Rob Portman (OH), Roger Wicker (MS), Todd Young (IN), Richard Shelby (AL), Thad Cochran (MS), Pat Roberts (KS), Shelly Moore Capito (WV), Roy Blunt (MO), Deb Fischer (NE), John Hoeven (ND), John Boozman (AR), and John Cornyn (TX). This group is largely made up of long serving senators and senators either in the GOP leadership or close to it. Mitch McConnell’s votes don’t highly correlate with this group but their votes represent the consensus position of the Republican leadership.

Another large group of Republicans include Jim Lankford (OK), Ron Johnson (WI), Chuck Grassley (IA), Jim Risch (ID), Mike Crapo (ID), and Richard Burr (NC). Their votes are distinct from leadership. Other small groups include Tom Cotton (AR) and David Perdue (GA) of legal immigration cuts fame, Mike Enzi (WY) and Ben Sasse (NE), and Jim Inhofe (OK), John Barrasso (WY), and John Thune (MT). Barrasso and Thune are in leadership but vote less often with the leadership consensus than others.

On the Democratic side, there are two large groups closely together. One contains Maggie Hassan (NH), Bob Casey (PA), Sheldon Whitehouse (RI), Bill Nelson (FL), Jeanne Shaheen (NH), Jack Reed (RI), Tammy Duckworth (IL), Gary Peters (MI), Debbie Stabenow (MI), Tim Kaine (VA), and Chuck Schumer (NY). The other group contains Ben Cardin (MD), Sherrod Brown (OH), Brian Schatz (HI), Martin Heinrich (NM), Tom Udall (NM), Tina Smith (MN), Chris Van Hollen (MD), Dick Durbin (IL), and Tammy Baldwin (WI). These groups are linked through Ben Cardin and Tim Kaine and Chuck Schumer. These senators are closest to the Democratic leadership.

A group liberal Democratic senators vote often with each and less with the leadership. This includes Ed Markey (MA), Bernie Sanders (VT), Kirsten Gillibrand (NY), Jeff Merkley (OR), Kamala Harris (CA), Elizabeth Warren (MA), and Cory Booker (NJ). Other distinct groups include Mark Warner (VA) and Amy Klobuchar (MN), Maria Cantwell (WA) and Michael Bennet (CO), Angus King (I-ME) and Tom Carper (DE), and Robert Menendez (NJ), Chris Murphy (CT), Jon Tester (MT), and Pat Leahy (VT).

The other half of the Senate have more individualistic voting patterns.

Zooming out, looking at basic correlation (50% or higher), unsurprisingly shows the two distinct parties.

Joe Manchin (WV) and Joe Donnelly (IN) serve as the bridge between the two parties. Manchin correlates with Donnelly and Republican Shelly Moore Capito (WV). Donnelly correlates with Heidi Heitkamp (ND) who then correlates with the blob of Democratic senators. The Republicans have more outliers from the party core. Ron Paul (KY), Mike Lee (UT), Jeff Flake (AZ), Johnny Isakson (GA), Lamar Alexander (TN), Mike Rounds (MT), Ted Cruz (TX), Susan Collins (ME), Cory Gardner (CO), and Lisa Murkowski (AK) break most often from the party and from each other. The Democrats are voting more cohesively this year.

The Des Moines Register came out with a poll this week showing that voters in Iowa are pretty much done with President Trump and more support Democrats for Congress than Republicans. However, that’s about all the good news in it.

Trump beat Hillary by almost ten points (51.2% to 41.8%) in 2016, higher than his margin in Texas. Voters loathe him now (35% Approve, 60% Disapprove). However, this does not translate to other Republicans in office. A majority approve of the job Governor Kim Reynolds is doing (51%-30%). Joni Ernst is hardly squealing as 48% approve of the job she is doing while just 38% disapprove. Charles Grassley has fossilized with majority approval as well (51%-40%). Even Mike Pence does well here (48%-42%), the first state he will campaign in if and when Trump is deposed from office. Hardly a rejection of Republicans, just a President many voters don’t view as the leader of his party.

Despite a third (33%) of voters not liking her, Kim Reynolds remains the best known (77%) and best liked (44%) candidate for Governor in a crowded field of nobodies (so far). Sure 49% prefer “someone new” over her (35%). However “someone new” does not equal a Democrat facing several million dollars of Koch brothers funded smears, something Tom Vilsack did not have to weather in his upset 1998 win in an open race.

Voters prefer Democrats over Republicans for Congress by a 40% to 33% margin. This is most pronounced in the first and third districts currently held by Republicans Rod Blum and David Young. However, the election is not for a year, 40% is far below a majority, and Democrats held leads in CD-3 and CD-4 at this point in 2013 and did not prevail the following year.

Despite voter energy helping Democrats beat expectations and prevail in New Jersey, Virginia, Alabama, and other special elections, voters are more turned off by politics (61%) than engaged (33%) since the 2016 election. Those who are more engaged are REALLY more engaged, but unless Democrats listen to the hopes and dreams of people tuning out, and offer an agenda relevant to their lives, 2018 will prove disappointing.

Democrat Doug Jones beat Roy Moore in Alabama last night 49.9% to 48.4%. As the maps below show, Jones did better than Hillary Clinton in many part of the state. Moore flipped many counties, particularly in the Mobile area.

This is all the more stunning because Democrats haven’t been able to compete in Alabama since 2006 where they have been blown out consistently in statewide races. Now that a Democrat has won, we can look at how a winning coalition in Alabama has evolved over time.

From 1998 to 2006, Democrats were involved in several close races. These include:

• 1998 Lieutenant Governor (49.6% Dem, 50.2% GOP)
• 1998 Attorney General (49.7% Dem, 50.2% GOP)
• 2002 Governor (49.0% Dem, 49.2% GOP)
• 2002 Lieutenant Governor (51.5% Dem, 46.8% GOP)
• 2006 Lieutenant Governor (50.6% Dem, 49.1% GOP)

In 2012, Roy Moore barely won election to the Alabama Supreme Court (48.1% Dem, 51.8% GOP). In averaging these six contests, the Democrats averaged 49.8% — almost the same as Jones’s 49.9%.

Geographically, though, Jones cut a new path to victory. Compared to this average, he outdid previous Democrats in the larger counties of Montgomery, Jefferson, Lee, Tuscaloosa, Mobile, Shelby, and Baldwin.

Moore, on the other hand, did much better in a sizeable number of small counties, particularly north of Jefferson and outside of Hunstsville, and in the southeastern part of the state.

This mirrors the national trend exacerbated in the 2016 election where Democrats perform well in highly populated parts of the country and poorly in small, white, rural counties with a small share of whites with college degrees. Doug Jones showed how Democrats can win in this environment in a Republican state with a large African-American population.

Oh, and we should never believe nonsense like this again when the media preemptively blames African-Americans for Democratic losses that have yet to even occur.

Hillary Clinton will weigh in next week with her take about “What Happened” in the 2016 election. Takes have abounded over the last 10 months, but the data released by the Voter Study Group gives some simple answers, namely that Hillary did a worse job holding on to Obama’s voters than Trump did Romney’s. It’s not all that interesting of a story without the insanity of last year (and this one). The narrative of a big Clinton win that came out of the debates and Access Hollywood tape was wrong and we need a better way to craft and evaluate election narratives to prevent similar delusions in the future.

The Voter Study Group is unique among election studies in that they re-interviewed people who they surveyed in 2011 and 2012. The self-reporting of the 2012 vote should be more accurate since respondents answered in the moment, not four years later like in the Cooperative Congressional Election Study (CCES). The electorate can be broken down into the following groups:

• People who voted in 2012 and 2016 (87.8% of the electorate)
• People who voted in 2016 and were eligible to vote in 2012, but said they did not vote (8.1% of the electorate)
• 18–21 year olds who voted in 2016 but were not eligible in 2012 (4.1% of the electorate based on an estimate from the CCES)

Among those who voted in 2012 and 2016, Obama beat Romney 50.5% to 44.9% (slightly higher than Obama’s 51%–47% overall margin in 2012 but a certain percentage of the electorate has died since 2012 and older voters went more Romney than Obama). The gap between Hillary and Trump with these voters narrowed considerably.

Dan Pfieffer articulated a familiar refrain many times on his podcast Keepin it 1600: “Who is the voter that voted for Obama and is voting for Trump or not voting?” He implied that these voters didn’t exist. Unfortunately, the 2012 Obama electorate is not synonymous with a progressive majority, and the “Never Trump” Republican didn’t exist outside of Republican consulting firms and DC green rooms. Hillary ultimately held on to smaller share of Obama voters (86.4%) than Trump did Romney voters (88.6%). She did not gain the Republican voters her campaign tried to woo.

An even bigger problem for Hillary was that among the Obama voters who voted against her, twice as many went to Trump than went to Jill Stein, Gary Johnson, Evan McMullin, and others on various state ballots. Among Romney voters who listened to what Mitt Romney said about Donald Trump, almost as many voted for a third party candidate as voted for Hillary. Those who went third party in 2012 split more to Trump than either Hillary or back to a third party candidate. Late campaign polls showed that this could happen, but this got lost in the Hillary win probability projections. If the “No Hillary” Obama voters split evenly between Trump and third party candidates, she could have won the four states she narrowly lost.

Another story that came out of 11/9 was the surge of new voters who came out for Trump. That didn’t happen either. Among 2016 voters who were eligible to vote in 2012 but didn’t, Trump and Hillary split almost evenly.

Among voters who were too young in 2012, Hillary won easily (from CCES since these voters were not eligible to participate in the Voter Study Group either).

More of those who voted in 2012 but did not participate in 2016 went for Obama than for Romney. This fits in with anecdotal evidence of Democrats dropping out of the electorate (among those alive and not voting in 2016. Pollsters have enough trouble interviewing living people to even think about surveying the dead).

The Voter Study Group clearly shows why Hillary came up short and did not get a high enough of a share of the national vote to win 270+ Electoral Votes — she was deeply unpopular. Overall, 42% rated her favorably while 56% viewed her unfavorably. Trump got a similar rating (44% favorable, 53% unfavorable) but was not any more unpopular than Hillary nationally despite running an insane campaign. Among those who voted third party in 2012 or 2016, or voted Obama-Trump or Romney-Clinton, Hillary was easily the most unpopular national politician, far surpassing Trump:

While he didn’t face the rigor of a national general election campaign and a billion dollars in Koch Brothers-funded attack ads, Bernie was far more liked than Hillary, Obama, or any leading Republican. Hence the song:

Lots of factors went into Hillary’s defeat. No single factor stands out. The race was always going to be close. Trump was going to hold almost all of Mitt Romney’s voters and Hillary’s unpopularity should have lowered expectations for her throughout the campaign among her supporters and in the media. On a brighter note, Trump in 2020 faces a large share of new young voters opposing him that will only grow as more of these voters participate and new young voters enter the process while elderly white voters who backed Trump “age out” of the electorate.

Every election cycle, the National Rifle Association (NRA) takes their members’ money and showers it on the Republican candidates Mitch McConnell and Paul Ryan tell them to. When the election turns out well for the Republicans, the NRA rushes out to take credit:

NRA demonstrated precisely why we and the gun owners we represent are the most powerful force in American politics. Thanks to the unique passion of NRA members and gun owners, we have an unparalleled power to mobilize voters in support of candidates who defend our rights, and against those who oppose them.

The NRA, with an assist from Bill Clinton, has carefully cultivated this narrative since 1994:

In 1994, Bill Clinton signed the Violent Crime Control and Law Enforcement Act into law, commonly known as the “Clinton crime bill.” The legislation contained a ban on hundreds of semi-automatic firearms. That year, NRA worked aggressively in the midterm elections to defeat lawmakers who supported the semi-automatic ban.

As a result of their support for gun control, Democrats lost control of Congress. Bill Clinton would later write in his autobiography that “The NRA had a great night. … The gun lobby claimed to have defeated 19 of the 24 members on their hit list. They did at least that much damage and could rightly claim to have made [Newt] Gingrich the House Speaker.”

Also in his book, Clinton identified the NRA as responsible for former Vice President Al Gore’s failed 2000 presidential campaign. In particular, he pointed to the NRA’s influence in Tennessee, Arkansas, West Virginia and several other states.

This story of 1994 and 2000 has been thoroughly debunked, but the NRA myth has recently returned with the two big Republican wins in 2014 and 2016. Digging into the campaign finance data, the NRA and its message has no impact in the outcome of federal races.

The NRA placed big bets on the 2016 elections. They spent over $30 million supporting Donald Trump and attacking Hillary Clinton. Guided by the Republican congressional leadership, they spent $21,728,979.20 on Republican congressional candidates and 86% of those dollars went to winning candidates. The bulk of their independent expenditure money went to attacking Democrats ($17.4 million, 79.9%) and in Senate races (79.6%, $17.3 million). They put the most money attacking Deborah Ross in North Carolina. However, in the races they attacked Democrats, they represented only a small share of the total amount of outside money spent in the races they chose.

Overall, groups spent $818 million dollars to influence contested congressional elections where Democrats faced Republicans (this analysis does not include uncontested races or ones in CA/LA/WA where Democrats faced Democrats or Republicans took on Republicans). In building a model to predict Democratic victory, using logistic regression, seven variables can predict 98.5% (396 out of 402) of the outcomes correctly. These include:

• Whether the Democrat was an incumbent or not
• The percentage of the vote Obama received in 2008
• The Democratic share of total candidate spending
• Whether the seat is open and represented by a Democrat
• The amount of independent expenditure money spent opposing the Democrat
• The amount of independent expenditure money spent supporting the Democrat
• The amount of independent expenditure money spent opposing the Republican

Independent expenditure money supporting the Republican was not significant in predicting whether Democrats won or lost in 2016. The $4.4 million the NRA spent promoting Republicans made no significant difference (this goes similarly for the $122 million other groups spent on this activity).

It takes a lot of money to make an impact with independent expenditure spending. In isolating the effect of each factor:

• Being an incumbent increases the probability of a Democrat winning by 4.2 percentage points.
• A percentage point increase in share of the vote Obama received in 2008 increases the probability of a Democrat winning by 0.7 percentage points.
• A percentage point increase in Democratic share of total candidate spending increases the probability of a Democrat winning by 0.04 percentage points.
• Spending a million dollars in independent expenditures opposing a Democrat decreases the probability of a Democrat winning by 1.0 percentage points.
• Spending a million dollars in independent expenditures supporting a Democrat increases the probability of a Democrat winning by 1.5 percentage points.
• Spending a million dollars in independent expenditures opposing a Republican increases the probability of a Democrat winning by 0.7 percentage points.

So when the NRA spent $5.6 million attacking Deborah Ross, we can expect that it reduced the probability of her winning by 5.8 percentage points. On the surface that seems significant (despite that she only had a .2% chance of winning heading into Election Day). However, is this a result of money spent by the NRA on the “gun rights” message or simply because the NRA spent money attacking a Democrat? Separating NRA spending from other independent expenditures shows no significant impact for the NRA’s money.

The same thing was true in the 2014 congressional midterms:

The 2014 model was a little different. Obama’s 2012 vote was more significant than his 2008 vote and independent expenditure money supporting Republicans was significant in predicting the probability of the Democrat winning while money boosting a Democrat was not. Isolating NRA spending, their message had no impact on the probability of the Democrat winning, just their money.

The NRA has a great marketing scam. They tell their members that they should give money to “protect the Second Amendment” and the NRA is the most powerful lobbying group in the country that single handedly wins election. This data tells a different story. The NRA’s money is only impactful in its being green. Mitch McConnell and Paul Ryan gleefully accept it because it helps Republicans win. That’s the extent of it. The “gun rights” message of the NRA has no significant impact and is not winning elections. NRA members might as well just give their money to support the Republican leadership in Washington and cut out the middle men in Northern Virginia. They are only subsidizing million dollar plus salaries of Wayne LaPierre and other executives and consultants for ineffective work.