Lies, Damned Lies and Partisanship

The latest Economist-YouGov national tracking poll includes a few great questions on public trust in government statistics. What’s better is they were previously asked in 2014, which allows for a comparison of views under the Obama and Trump administrations.

Key takeaway: while Republican views on government statistics have evolved – they now have considerably more confidence in the numbers than they had in 2014 – Democratic views have not shifted much at all.

Percentage who say half or more of the statistics reported by the government are reliable and accurate:

May 2014 Mar 2017 Change
Republicans 32% 58% +26
Independents 50% 56% +6
Democrats 74% 77% +3
Total 50% 64% +14

SOURCE: Economist-YouGov Polls

You see the same thing when the question gets more specific: is the unemployment rate being accurately reported?

Percentage who say there are “More unemployed people” than shown in the numbers released by the government in the most recent employment report:

May 2014 Mar 2017 Change
Republicans 76% 48% -28
Independents 64% 46% -18
Democrats 41% 39% -2
Total 59% 44% -15

SOURCE: Economist-YouGov Polls

There are similar effects to varying degrees for CBO data on the number of uninsured, US Census data on the population, and NASA/NOAA data on temperature changes.

It’s possible that Democrats’ trust in Trump administration will deteriorate over time. It might even be surprising if that didn’t happen. However, it’s also plausible is that Democrats are, as a starting point, much more likely to be confident in official government information.

In fact, pushing a “both sides do it” frame on this seems forced – and worse, for a blog post, boring. The more interesting theorizing has to do with the question of why Republicans tend to be more epistemologically flexible when it comes to official statistics. My hypothesis is that these people have been forced to choose between political or religious realities and the word of “the BLS” or “the media” or “climate scientists” or “crowd scientists”  in ways most Democrats simply have not (or at least have not had to do so on such a regular basis). As a result, the default position on official and expert data shifts from credulity to something purely conditional. Democrats are surely victim to this in certain ways. But it seems to me that they are not forced to choose between the “official” line and their own moral line nearly as often as Republicans are.

Lies, Damned Lies and Partisanship

The New Quinnipiac Poll: Either Noise… or Pretty Worrying for Trump

Quinnipiac (KWIN-uh-pe-ack) University has a new poll out, which is mainly interesting for some erosion in support among Trump’s base, as their poll release points out.

Quinnipiac always breaks out the information for whites with and without a college degree (something really all pollsters should do in the Trump era), which allows for a look at the trend lines since January. This shows that their previous polls had been relatively consistent among both groups — non-college whites gave Trump net ratings of around +20 and college whites around -15.

This week, whites with a degree haven’t budged, but his support among non-college whites has fallen significantly to a net +7, from a high in early March of +26. This is the first Quinnipiac Poll to come out since the GOP released its Obamacare replacement bill (on March 6th). Importantly, it’s not alone in showing a softening in Trump’s numbers during March, at least in the toplines.

Using “net” approval, percent approve minus percent disapprove, has its drawbacks – the measure is, basically by definition, noisier than just “approve” or “disapprove”. But it’s useful for comparisons across groups, and the changes are eye-catching however you look at it. Since February, Trump had been consistently in the 56-60% range for approval with non-college whites, and he is now at 50%. File under Number To Watch.

In addition, Trump’s support among Republicans declines from 91% to 81%. Disapproval is up to 14% from 5%. This may seem like a natural decline, or like it’s still quite high, but context is important here. George W. Bush was in the 80s and 90s among Republicans for most of his presidency – amidst numerous crises and big declines among Democrats and independents. He only fell consistently below Trump’s current level after his second midterm election in 2006, with the combined pressures of Katrina, Iraq and eventually economic crisis. And even at Barack Obama’s lowest points, he really slumped into the mid-70s among Democrats. In other words: maybe this is just noise, or else Trump hitting his floor with Republicans – but I don’t think “still strong with the base” is the correct frame if he’s really hovering around 80%.

The New Quinnipiac Poll: Either Noise… or Pretty Worrying for Trump

The Age Gap in the Age of Trump

Ron Brownstein has a new piece in The Atlantic interrogating Donald Trump’s support—or lack thereof—and what it might mean for Democrats. It contains some eye-popping numbers, such as the huge differential swings among white Millennials with and without college degrees. Brownstein also notes Trump’s very low approval ratings among 18 to 29 year olds, which have ranged between the low-20s and the low-30s depending on the pollster.

This raised a secondary question for me about age cohorts in the Age of Trump: namely, is this anything new? In the background of these thoughts is the common misconception that young people have always tended to vote more Democratic than the rest of the population. Actually, while the age gap existed during Hippie Era, it essentially went into hibernation until George W. Bush’s 2004 reelection and then exploded during the Obama years.

If anything, the relative advantage Democrats have held with younger generations is anomalous until proven otherwise, so it is notable (if not necessarily surprising) if it endures into the Age of Trump. To that point, I looked at the age variation in approval ratings across recent presidencies.


As I mentioned, the age gap is a recent development. If you take the average approval ratings from their entire presidencies (as Gallup has done here), there is virtually no difference between levels of support from the youngest and oldest voters. In the Obama years, this changed, and on average his approval rating was 16 points higher among 18-29 year olds than it was among Americans over the age of 65. Based on recent polling of Donald Trump, he may indeed be headed for an even bigger gap: according to the latest Economist/YouGov Poll, he is getting 28% from 18-29 year olds and 52% from over-65s—a yawning gap of 24 points.

However, a bit more context is probably necessary, because averages can be misleading when seeking out trends. The young/old gap in approval actually varied quite a lot during Obama’s presidency, and if you take only his last year in office, the age gap was much larger—24 points, in fact, exactly the same as under Trump (but reversed, with positive ratings higher among young people).


So, in one sense, the age gap in the Age of Trump is picking up from where it left off under Obama. On the other hand, while Trump has gained only 9 points on Obama’s 2016 average among the oldest age cohort, he trails Obama’s average among young people by a massive 39 points, and even trails Bush 43’s average by 18 points. Millennials continue to vote at much lower rates than older voters, so Republicans have generally had the better end of the deal when it comes to the age gap. Whether that continues to be the case as Millennials grow from 30% to 45% of the electorate over the next 8 years… well, we’ll just have to wait and see.

The Age Gap in the Age of Trump

What’s a winnable seat?

Increasingly, Democrats are turning their attention to the congressional race in GA-06, the district of Tom Price, Donald Trump’s new Health & Human Services Secretary. The first round of this election is April 18th*,  and it will inevitably come to be seen as a test case for how transferable the energy seen in ongoing anti-Trump protests is to defeating Republicans down ballot. Progressives have already raised around $600,000 to win the seat (Donate here).

But there’s another twist. While Tom Price won his race there decisively (62-38), it’s one of a handful of traditionally red, suburban districts that swung dramatically towards the Democratic presidential ticket in 2016. For example, while Mitt Romney won GA-06 by 23 points in 2012 (61-38), according to Daily Kos Elections, Donald Trump won by less than 2  points (48-47). Arguably, educated, suburban and often increasingly diverse districts like these could be key to Democrats future if they are unable to separate Trump from the less educated, whiter base that lifted him to victory last November.

This prompts a question: will Democrats be able to win these seats without appeasing their Democrat-curious Romney-voting electorates by tacking sharply to “the center”?  My instinct is yes. It deserves a closer look, but the chart below comes to mind. It’s from the excellent book Democracy for Realists by Christopher Achen and Larry Bartels (buy it, read it).

SOURCE: Democracy for Realists by Christopher Achen and Larry Bartels

The illustration shows how the ideology of constituents** compares to the ideology of their representatives, based on roll call votes.

The broader question of why voters vote the way they do is obviously a deep and complex one, but it’s worth making a simple point. There are many reasons voters choose one candidate over another, from incumbency and name recognition, to national trends or simply candidate quality and missteps. Could suburban Georgia voters reject a candidate who comes to be seen as radically left-wing? Of course. But, based on the above, it seems likely that a good candidate in good conditions is just that. There are a lot of “moderate” districts that end up with representatives that vote in a very conservative way (in a large part because they are Republicans) and other, equally moderate districts with representatives that toe the party line in the other direction.

Democrats have real opportunities here.

* However, Georgia uses a runoff system, and if no candidate reaches 50% there is another election on June 20th for the top two contenders.
** Based on the massive 52,000-respondent Cooperative Congressional Election Study (CCES) conducted by YouGov in 2012. The ideology measure combines response from a range of issue-based questions, from abortion to Obamacare.
What’s a winnable seat?

Donald Trump’s Approval Rating in Context

Just how bad is Donald Trump’s approval rating, historically speaking?


It has already been pointed out that the current President rose to majority disapproval far more quickly than any other president, but I wanted to take a closer look at how his current standing compares to the record of other Presidents. Clearly, it is very early (actually, that’s sort of the point) and far too early to draw any serious conclusions, but still, it’s an interesting question from a historical perspective.

At the top of this post is the Gallup net-approval rating – that’s percent approval minus percent disapproval – for every President going back to Harry S. Truman, the first for whom there’s early-Administration polling. Trump’s net-approval is -10 today, up slightly from -12 yesterday. Clearly, this is low. The average rating for presidents during this period is +49, ranging from +23 for Bill Clinton to +75 for a post-JFK assassination Lyndon Johnson (though Truman, who was busy wrapping up a World War, scored +84 in his first poll on day 48).

In fact, three presidents – Eisenhower, Ford and Kennedy – never scored as low as Trump is now, for the entirety of their time in the White House. However, two of them did not even serve four years in office, restricting their ability to disappoint the country.


For the others, it took much longer to reach this far into the presidential doldrums. The closest second is Bill Clinton, who reached -13 in a poll conducted during June of his first term. The earliest stages of Clinton’s presidency saw the Travelgate controversy draped over an already scandal-clad President, who won election with only 43% of the vote. For other presidents it took much longer: next in line was Truman, who dropped below -12 a full year and a half into his tenure. On the other end of the scale are George W. Bush and Richard Nixon, both of whom wouldn’t fall below Trump’s current level until their second terms.


The Bush and Nixon examples reintroduce some nuance into our assessment of Trump’s current standing. George W. Bush took longer to hit this level than any other modern president, but ended his time at the White House with lower ratings than any other resident who lasted a full two terms. Bill Clinton, though he got off to a uniquely rough start, ended up with some of the highest approval ratings of any two-termer. This point is captured compellingly in a recent piece by Princeton historian Julian Zelizer, who recounts Ronald Reagan’s answer to public protest, international disorder and low approval ratings: decisive re-election.


All the same, it’s hard to ignore how Trump in not even three weeks has reached depths that usually inspire media folklore of presidencies trawling across the ocean floor, searching for direction. Here are some comparable periods for recent Presidents, who faced some of the same partisan pressures that likely delimit both the floor and the ceiling of the modern-day approval rating:

Bill Clinton Paula Jones suit, Whitewater hearings, death of Hillarycare Sep. 1994 -10
George W. Bush  Hurricane Katrina and Aftermath Sep. 2005 -9
Barack Obama Debt-ceiling crisis, S&P downgrade Aug. 2011 -11
Donald Trump Two weeks as President Feb. 2017 -10

In addition, were Trump’s ratings to remain in their current territory for the next 20-odd months, history suggests his party would be headed for a solid rebuke at the Midterm elections. Every president with an approval rating under 55% has lost House seats in the midterms, and losses for presidents under 50% range from 11 seats (1978) to 63 (2010). Notably, Democrats need to gain only 24 seats to retake control of the House of Representatives, and the only circumstance where a president lost fewer than that with comparable ratings was 2014, when an already diminished Democratic Party lost 13 seats six years into Obama’s presidency. The average seat loss for a president in the 39% to 46% approval range is 39. Trump is at 43%.


Of course, nothing about Donald Trump or his campaign for president has fit into the template of the average President. And in 2016, House Democrats lost double the number of districts won by Hillary Clinton than Republicans who lost Trump districts, and it’s possible they will repeat the performance in 2018. But a repeat performance would require Trump to break dramatically with history, once again.

Donald Trump’s Approval Rating in Context

Trump’s Muslim ban and public opinion

Date Pollster Question Response options Pro-ban (%) Anti-ban (%)
Jan. 30-31 PPP Do you support or oppose Donald Trump’s executive order banning refugees and citizens of certain countries from entering the United States? Support-Oppose 47 49
Jan. 31-Feb. 1 YouGov President Trump recently signed an executive order banning travel for people from seven Muslim-majority countries – Iran, Iraq, Syria, Sudan, Libya, Yemen and Somalia – for 90 days, and suspending the admission of refugees for 120 days. Do you approve or disapprove of this ban? Approve-Disapprove-Not sure 48 44
Jan. 30-31 Ipsos Do you agree or disagree with the Executive Order that President Trump signed blocking refugees and banning people from seven Muslim majority countries from entering the U.S.? Agree-Disagree-Don’t know 49 41

Issue polling is a dicey business, for a few reasons.

For one, policies by definition deliver some kind of public “good” to some constituency, at least superficially. Focus on that, and you can probably get a positive response. At the same time, there are always trade-offs, and it can be very difficult, if not impossible, to capture those trade-offs in a meaningful way (and sometimes pollsters don’t even try). And because most voters don’t care about The Good Or Maybe Bad Policy nearly as much as you do, they’re going to defer to you on the pros and cons. Second, there’s an important distinction between testing a policy as a political win for a party or politician and testing it as a piece of independent political communication that is often lost. Testing one side of that divide or the other – or some combination – is appropriate in different contexts, but they illustrate very different things.

For some of these reasons, I was worried about polling on the #MuslimBan, which involves components that have polled well before, but is already pretty clearly tied to a popularity-challenged president. Anyway, it looks like I was wrong!

Polling on the policy is, so far, pretty consistent, despite some variation in question and response options. The public is fairly evenly divided, with supporters slightly outnumbering opponents in two out of three polls and the third essentially evenly split. All of the polls mention Donald Trump by name, which I think is correct, because abstracting Trump out of this is beside the point. There is a possibility that omitting detail about the countries involved is responsible for wider opposition in PPP’s case, but that could have other explanations, up to and including regular sampling error. At this stage it’s not an especially popular policy or an especially unpopular one (all the more reason to focus on the fact that it’s an immoral and ineffective one).


Below is a just-published Gallup poll, which paints a much more negative picture of public opinion and the ban. It’s not obvious to me what’s going on here, but it does introduce the possibility of a modal effect – perhaps people are eager to tell phone interviewers they dislike the ban, but more non-committal when being polled online. As noted yesterday, Gallup has also been more negative for Trump on regular job approval.


Trump’s Muslim ban and public opinion

The Era of Strong Feelings

This week’s installment of the weekly Economist/YouGov Poll was published today, adding another approval rating to the mix – 43% approve to 44% disapprove. Disapprove rose by 9 points from last week, while disapproval also rose by 2. It’s still very early and there is much going on, so it’s difficult to draw too many conclusions about any individual poll at this point. However, in making sense of the relatively low disapproval number (when compared to Obama, who was in mid- to high-50s for years) and the loud, proud protests breaking out across the country, the “Strongly disapprove” number does stand out.

Roughly a third (33%) of the country already “strongly” disapproves of Donald Trump, making up, you might say, the floor of anti-Trump sentiment. Obama didn’t reach this level of passionate opposition until December 2009, following months of backlash and Tea Party protests over TARP, taxes and the impending “government takeover” of healthcare. Afterwards, those numbers never really went below that level again.


What this means for Trump’s presidency is uncertain. Obama, of course, lost out in the midterms in 2010, but went on to a convincing reelection in 2013. But when it comes to the formation of a sizable, hardened opposition, Trump has received no honeymoon.

Sidebar: There’s now quite  a spread on Trump: Politico/Morning Consult have the approve-disapprove split at 49-41 (albeit pre-ban) and Gallup have 52-43. One wonders about a modal effect – are respondents online more willing to be honest about positive feelings towards Trump? – but YouGov doesn’t really confirm that. However, one of the under-appreciated differences between the online and phone “modes” is simply the options provided; YouGov and Morning Consult display a “don’t know” option while Gallup’s is volunteered. All three show undecideds in decline, but one additional possibility for Gallup’s higher disapproval number is people who haven’t really formed an opinion are reverting to their general feelings about Trump, even independent of his presidential performance.

The Era of Strong Feelings