Thanks to a name pulled out of a film canister, the race for control of the Virginia House of Delegates race has probably been settled. Republican incumbent David Yancey will likely hold onto his seat in the Virginia House of Delegates, defeating challenger Shelley Simonds and giving Republicans a 51-49 seat edge.
But Yancey does have something in common with 14 of the 15 Republicans who were not so lucky on election night (or thereafter): he’s a white dude. Remarkably, just 3 of the incoming Democrats can say the same for themselves. This fact, on top of the generational trends seen in the electorate and the evidence that there is an unprecedented surge in women – the vast majority of them Democrats – running for the U.S. House of Representatives, got me wondering just how much of a demographic change the Virginia House of Delegates will undergo when the new class is sworn in on January 10th.
I took a look under the hood, and the numbers really are pretty striking. Below is a chart showing how the make-up of the Virginia House of Delegates will change following the results of the election on November 7th, 2017 (all the numbers used here assume Yancey will keep his seat).
There are 100 delegates, so each percentage point shift technically represents one delegate, but as you will see below it is useful to keep the percentages for some further comparisons. First, however, let’s review the overall damage.
Overnight, the share of white men in the Virginia House of Delegates will fall by 12 points, from 71 percent to 59 percent. Meanwhile the share of women in Virginia’s lower house nearly doubles, from 17 percent to 30 percent, as does the share of delegates under the age of 40, from 10 percent to 19 percent. Delegates from communities of color increase more modestly, from 19 percent to 23 percent.
These are some pretty big changes, though this composition falls notably short of being fully “representative,” which seems like a decent benchmark. Below I’ve copied the chart from above, but included some dotted lines to show what the actual demographic composition of the Virginia population is (according to recent Census estimates). Roughly 31 percent of Virginia’s population are white men. Fifty-one percent are women. Nearly 53 percent are under the age of 40, and 38 percent come from communities of color*.
This may not be the best way to look at the numbers – it certainly isn’t the only way. After all, the election was big news because virtually all of the new delegates areDemocrats. So it seems worth looking at the shifts (or lack thereof) within parties as well as overall.
For Republicans, the changes are notable in how small they are. All but one of the defeated Republicans were white men – and, of the three Republicans newly elected to replace fellow Republicans who are retiring, one was a woman** – but the share of the remaining Republicans who are white men barely budges from 89 percent to 88 percent, because the outgoing Republicans are fairly representative of the group as a whole. The share of women in the Republican caucus does increase from 6 percent to 10 percent, though that’s due to the addition of one woman to the existing four, along with the decline in the number of Republicans (from 64 to 51) overall.
Among the Democrats, the changes are more dramatic. Nearly all of the new Democrats are women, and as a result, the number of Democratic women in the caucus virtually doubles from 13 to 25. In percentage terms, that means the share of women in the Democratic caucus of the Virginia House of Delegates will be 51 percent – exactly representative of Virginia’s population. The influx of new Democratic delegates is slightly whiter as a group than the incumbents so, overall, the share of Democrats from communities of color falls from 47 percent to 43 percent even as the share of white men in the caucus falls from 35 percent to 31 percent. The biggest change, however, is generational. The share of Democrats who are under 40 years old more than doubles from just 12 percent to 27 percent.
One way to interpret these trends is to note that in nearly every case the Democratic caucus comes much closer to looking like the population of Virginia, particularly in the case of gender and age.
What does it all mean – beyond Virginia? Maybe nothing.
But, as my headline suggests, we might look at this as a harbinger of what to expect in 2018, when the entire House of Representatives and many Senate seats are up for re-election. I have not seen the numbers broken down by age or race, but, as I mentioned above, there is a stunning, unprecedented number of Democratic women running for office next year. The Virginia story may be reflective of demographic trends occurring within the broader Democratic Party, both among Democratic voters and among Democratic candidates – trends that have been virtually absent within the Republican Party, and which have in some ways been hidden from view because it’s been Republicans winning the majority of elections since 2008.
As it stands, there are 62 Democratic women in the House of Representatives. That’s about 74 percent of all the women in the chamber, but still just 32 percent of the Democratic caucus – far short of representative of the country. Virtually all (around 95 percent) of Democrats in the House are 40 or older; most are over the age of 60.
I would put some money on that changing quite a bit next year.***
* For this analysis when I say “persons/communities of color” I include Hispanic/Latinx people who may otherwise identify as “white” when responding to the Census.
*** As with the Virginia House of Delegates, if Democrats win a lot of U.S. House seats their caucus may actually get whiter even as the overall House gets more diverse. For a number of reasons (the Voting Rights Act, gerrymandering and geography I suspect are the biggest ones), some of the safest Democratic seats are held by persons of color. So I would guess the new cohort of Democrats – who will naturally win more competitive seats – will be more diverse than the Republicans they replace, but whiter as a group than the current class of Democrats. That said (once again, similar to the Virginia case) this will technically make the racial mix of the Democratic caucus more representative of the nation as a whole, because 43 percent of the current Democratic caucus are racial or ethnic minorities, versus 39 percent of all Americans.
Tuesday, November 7th, 2017, was a big night for Democrats. Democrats won a lot of races they were not expected to win, especially in the well-educated suburbs that were supposed to make the difference for Hillary Clinton last year, but didn’t.
One thing that stuck out to me, though, was what the exit polls told us about young voters. Here’s what I tweeted that night:
Whoa. Let’s not lose sight of the youth vote here.
Later in the week I put together a more detailed table tracking the youth vote, displayed below.
The additional context demonstrates a couple of things. For one, McAuliffe’s relatively poor performance with young voters in 18-29 may have been a fluke, or simply polling error (exit polls are, after all, polls, and they have error). Yet the fact still remains: Northam did extremely well with young voters, capturing a larger percentage of the 18-29 year-old vote than any statewide candidate in recent memory, with the exception of Mark Warner. Ex-Gov. Warner won 71 percent of the youth vote in 2008, but that was while he was winning 65 percent of the vote across all ages against ex-Gov. Jim Gilmore.
Northam also did particularly well among the 30-44 crowd, who vote at higher rates and make up a much larger share of the electorate. One thing to note here that will become relevant again lower down: nearly half of 30-44s are now Millennials, the oldest of whom are now around 35 or 36 years old. Meanwhile, Northam did about average with voters over 45, and that’s key – it was the age gap between young and old that was especially unique.
I compiled a similar dataset for New Jersey, and the same picture emerged. In Murphy’s case, he really was the strongest candidate since at least 2004 among voters aged 18 to 29 (caveat: exit polls do not appear to have been conducted for Booker’s elections in 2013 or 2014, or for Lautenberg’s in 2008). Murphy won “only” 56 percent statewide, but took a massive 73 percent among 18-29 year-olds and 63 percent among 30-44 year-olds
This is one reason why talk of Northam (or Murphy) reassembling either the Clinton coalition, or the Obama coalition, strikes me a little off base. What happened this November is something relatively new, at least generationally.
In fact, Northam’s coalition first reminded me of an election that took place outside of Virginia – outside of the United States, actually.
The United Kingdom
Earlier this year, I noticed something funny about the three most recent UK-wide elections – the 2015 General Election, the 2016 EU Referendum, and the 2017 General Election.
I know there’s more to this story, but it’s pretty striking.
Age polarization exploded for the Brexit vote in 2016. Then, in the 2017 general elections, voters split along age lines again – and in a way that looked a lot more like the EU referendum than the previous parliamentary election.
I lived in London from 2008 to 2015, and I think (hope) I can provide a bit of anecdotal context for this. First of all, most young people in Britain – certainly those born in the late-1980s or after – literally can’t remember a Britain that wasn’t part of the modern EU. The last time the European question was officially on the ballot was 1975, a referendum on joining the European “common market,” a precursor of the Union (which was founded in 1993). The 2016 referendum campaign was about a lot of things, but the defining issue was immigration, with the “Leave” camp arguing for a whole lot less of it. More generally though, the debate about In-versus-Out became a proxy for a generational debate. The most powerful argument in favor of Leave was not really about policy, but a feeling of nostalgia for a previous version of Great Britain – one that no UK Millennial has ever experienced. But Remain ran an uninspiring campaign and its victory seemed inevitable to most people; turnout, particularly among young people, was dismal (sound familiar?).
The margin for “Remain” among young people, however, was gargantuan. When they effectively got a second chance to vote against Brexit – in 2017, by voting against the Tories who brought it about (and to vote for Jeremy Corbyn, who really had great appeal among young Britons) – the Remain margin returned, but, for the first time in a generation, young people actually turned out, and Labour massively outperformed expectations.
Trump’s approval is down a little across all age groups since then, but the general point I was making holds up. The gap in approval for Trump continues to look like an inverted version of the gap that prevailed for Obama, in a contrast with George W. Bush and Bill Clinton. A notable aside: Trump’s approval rating is much higher overall than Bush’s at the end of two terms, a failed war, and a trashed economy – but that’s only because of older voters; with younger voters, 2017 Trump is already about level with 2008 Bush.
Generally I think there is a belief out there that younger voters have always leaned further to the left than the rest of the county, only to grow more conservative as they grow older, marry, move to the suburbs, start paying taxes. As the graphic above illustrates, this was not generally the case for Clinton or Bush 43 in terms of average job approval.
Presidential exit polls over the past 40 or so years allow us to look even further back. In the illustration below I’ve taken the percentage support for the Democratic presidential candidate among each major age group and subtracted from that the national margin – this gives us the Democratic “lean” of each age cohort, essentially controlling for any big swings that affected every generation (as in the Mark Warner case above).
While 18-29 year-old age cohort, shown in turquoise, egregiously rejected Nixon in 1972, it had an average Democratic lean of less than three points from Jimmy Carter until Dubya’s re-election. Then, in 2008, the number shot up to a 14-point Democratic lean, then 10 points, then 9 points.
The trend isn’t limited to elections with Barack Obama – or Trump – on the ballot. Over the same 1976-2006 period the average Democratic lean of 18-29 year-olds in House elections was even smaller, just 2 points, and then it leapt to 9 points in 2008.
The House numbers actually suggest the trend slightly predates Obama, with a generational lean emerging during the Bush years.
Notable in both cases, if not as visually dramatic, is the uptick in Democratic lean among the 30-44 year-olds. Part of this may be Gen-Xers becoming more liberal, but the other shift to consider is that the Millennials who were in their mid- to late-20s when Obama was first elected are now in their mid- to late-30s. This suggests Obama-era Millennials are carrying their left-leaning voting habits into later adulthood, consistent with survey data from Pew Research that finds Democratic identification among the generation has hardly budged since 2008.
That older Millennials have remained about as Democrat-friendly as they were when they started voting is also consistent with research into generational voting habits over the past 60 years. Data scientist Yair Ghitza and statistician Andrew Gelman designed a model of presidential vote preferences and concluded that voting habits are formed in a cumulative process of evaluation of presidential performance over a voter’s lifetime [definitely click on this link to the NYT Upshot write-up of the research for a fantastic interactive data visualization]. However, they also found that events at all ages are not equally important – in fact, events that take place at age 18 are three times more potent in shaping later preferences than events that take place at age 40. People who entered their formative years under Kennedy, LBJ and Nixon remained relatively Democratic for years after, while those who did so under Carter and Reagan developed a lasting Republican lean.
The first Millennials turned 18 around 2000: this is a cohort of voters whose formative years took place during the economic boom-times of 2nd-term Bill Clinton, war and financial destruction under George W. Bush, and/or the Obama years. (Much, if not most of the upswing in Obama’s approval rating in 2016 came from Millennials. He left office with an insane 77% approval rating among Millennials – that’s even higher than when he first took office in 2008).
To many people this might feel like a waste of words. Didn’t we already know Millennials are liberals? The problem for Democrats isn’t that Millennials vote Republicans, it’s that they don’t vote at all.
There’s nothing remarkable about the rise of Millennials as a voting group per se. For a number of reasons – higher income, stable address, practice – previous generations voted more as they grew older, and most of the same factors apply to Millennials as well.
And, unlike in previous election cycles, we are further forward in time. By Pew’s definition (it is by no means the only one) the median Millennial was about 19 in 2008, and so s/he will be about 29 in 2018, a very different stage in life – and voting behavior.
A good illustration of this is from a Wonkblog article using data from Political Data Inc., a California firm. The curve shows turnout rates by age, and what you get is a spike for 18-year-olds and then a gradual, linear increase until turnout plateaus around age 70.
We can use this to see what’s significant about this moment in time for Millennials in terms of their electoral strength. Below, I’ve annotated the graphic to show how Millennials and Baby Boomers have progressed along the curve with each 2-year election cycle (the green box represents 2012).
In other words, Millennials are entering a phase where the whole generation, across all ages, gets more likely to vote each passing year. Boomers and their elders, meanwhile, are declining in numbers (sad, but true) and, in many cases, declining or plateauing in average turnout rates.
Again, this isn’t really unique to these particular cohorts, though I’ve used them as an example. Charles Franklin, who now runs the Marquette Law Poll and who is great on Twitter, produced a chart similar to the one above back in 2008, using 2000 and 2004 data. He went a step further and created a curve that shows the over- and under-representation of age groups in those two elections.
Older voters are “over-represented” in the sense that they make up a larger share of actual voters than they do of eligible voters (in 2016, for example, Millennials and Boomers made up the same share of eligible voters for the first time; but many more Boomers than Millennials actually voted). Franklin found that, for the two elections under consideration, a voter stops being under-represented in the electorate at about age 40 – which is the age of some of the oldest Millennials in around 2022.
Up until now, most of the benefits for Democrats of huge margins among younger voters have been cancelled out by the fact that they’ve lost a little ground among older voters. But due to their higher turnout rates, a “little” actually means a lot. Below is the “Democratic lean” graphic from earlier on, but (very roughly) weighted by average turnout for each age group. Immediately things even out.
It’s not obvious that the Millennial Reckoning will come in 2018, or even 2020. That said, based on data from past elections, it’s when a cohort starts to nudge up against their fourth or fifth presidential election cycle that they start seeing real improvements in turnout rates – and that’s about where Millennials are now. Looking just at how Boomers and Gen-Xers behaved around a similar time, we’d expect Millennial turnout to increase by around 5 or 6 points between 2016 and 2020.
According to one early estimate by Tufts University, turnout among 18-29 year olds increased by 8 points in Virginia and and was steady in New Jersey compared to the last gubernatorial election. Overall, exit polls suggest the 2017 electorate was slightly younger in both Virginia and New Jersey than the 2013 electorate.
Don’t Call it a Realignment
Sean McElwee argued in in Vice recently that the Democratic Party need to tack left on racial and social justice issues to court Millennials. Republican pollster Kristen Soltis Anderson made a case in 2015 that the GOP could, and should, work to win over this rising generation by finding conservative ideas Millennials care about.
This is a reminder that, even if Millennials’ preferences don’t change much, the parties might. It’s part of why political “realignments,” in which the balance of power between the country’s parties alters permanently are so rare.
Another reason is that shifts in one direction with certain groups are often outdone by shifts in the other direction, as was partly the case with young and old voters in the early 2000s. It may be the case that Trump makes up enough ground among older voters to avoid a generational wipeout, but that’s no given, especially with an approval rating now well below 50% among older voters, too.
We haven’t yet seen what it looks like when Millennials come into their own as a voting bloc, and it will almost certainly happen. The only question is, will it happen gradually, or suddenly?
Is there anything going on in American politics that might spur an unusual lurch forward among the nation’s youngest, most diverse and most educated generation? We’ll just have to wait and see.
Below is a list of mailers for Republican campaigns and GOTV operations I’ve come across online.
Notice the vast majority are from official arms of the Republican Party – the NRSC, the state parties of New York and Virginia. I’ve noted where the mailers were delivered and included links to local news coverage where have been called out for fearmongering and race-baiting. Republicans appear to have decided, based on some combination of Trump’s example and independent research I’d guess, that this is what they need to do to win. The only way to stop it is going to be showing them it won’t work.
This is a short post just to give a home to a few charts I made out of curiosity. YouGov* has quietly been carrying out a pretty huge project that may be interesting to future historians and social scientists. Heck, it’s pretty interesting now.
What they’ve done is included nearly every one of the president’s tweets since February 4th in an online survey (YouGov is an online-only research company) and asked people to rate them on a scale of Terrible to Great. This is recorded as a score from -2 to 2, then multiplied by 100, so there is a theoretical maximum of 200 (if every respondent said the tweet was Great). More detail can be found on their micro site here, where they have every tweet with its overall rating as well as its rating by party identification.
Hypothetically, a dataset like this could be used to see whether Trump’s bad tweets (or his good ones) have a measurable effect on other measures of his performance, like Job Approval. I haven’t yet made any comparison like that, but for now I’ve included some charts from this data set below. To be clear, I just scraped this from their site – I didn’t have any direct line. Also note there are some gaps in the data that I can’t explain; a week in September is missing; for roughly a month in June-July I could only get some tweets that were in the top or bottom 80% because the site wouldn’t load everything when sorting by date; there may be other quirks I’ve not noticed. But it’s a large dataset nevertheless, including 1,304 tweets since February 4th, 2017.
I’ll let the charts speak for themselves, but to make a couple of notes from the toplines: most of his tweets, 61% to be exact, are sub-zero tweets, meaning they received more bad ratings than good. The median tweet receives a rating of -15. The bottom 25% of tweets receive a rating of -40 and lower. Democrats are more likely to give his tweets a positive rating (55 tweets, or 4%) than Republicans are to give him a negative rating (1 tweet, or 0.08%). That one tweet was Trump’s Worst Tweet Ever, according to the Index. It was a tweet criticizing Mark Cuban.
(Note, the lines are created using LOESS regression).
Out of interest, I also looked at volume of tweets. It seems the president has been tweeting more and more since the spring. He was tweeting about 4 tweets a day in March; more recently, he’s been tweeting about 8-10 times a day. ¯\_(ツ)_/¯
*Disclosure, I used to work for them, though before this project was underway.
Ten days after Hurricane Maria made landfall, millions in Puerto Rico are still without electricity, gas, cash or running water (make a donation to help out here)—but this is 2017, so obviously we’re talking about the president’s bad tweets.
The meta-narrative here is that Trump is on the defensive against an emerging narrative that the administration has been tragically slow to respond to Maria, the third major hurricane to make U.S. landfall this season. In fact, the specific tweet that got me thinking about this post was actually from Press Secretary Sarah Huckabee, pushing back on a report published by the Washington Post overnight.
WP story on PR is false. @POTUS gave military & first responders 100% support & has been updated daily on efforts by Gen Kelly and @fema
The report itself is a truly damning, and depressingly plausible, account of Trump’s initial handling of Maria. It describes how Trump became alerted to the true urgency of the crisis in Puerto Rico only after watching television from his golf club in Bedminster, NJ, already several days after Maria made landfall and knocked out power for the whole island.
Though the obvious answer to “should we believe Sarah Sanders that Trump has been attentive to the crisis?” is “no” we are actually in a fairly unique position, historically speaking, to understand the president’s priorities and state of mind over this period. Because of his bad tweets.
So, as it’s a beautiful Saturday in NYC, I cracked open Excel and coded five weeks of the president’s tweets and retweets based on their subject-matter. The data is displayed in the charts below, with tweets about the three recent hurricanes—Harvey, Irma and Maria—highlighted. I also note the dates each hurricane made landfall in Texas, Florida and Puerto Rico, respectively, and the dates Trump visited each hurricane site.
Here’s the same data, but with each day’s topics shown as a percentage of all topics Trump tweeted about that day.
It’s pretty obvious which events really gained the president’s attention. In terms of hurricanes, Trump tweeted dozens of warnings and, later, supportive messages for Harvey. Similarly, Irma dominated his output when it seemed poised for a direct hit on southern Florida (it also hit PR several days earlier—killing three—thought only one Irma tweet mentioned the island). Maria has only come into the foreground the past couple of days—more than a week after landfall, and after coverage had already turned negative on the administration’s handling of a growing crisis. And while Trump visited both Texas and Florida within 5 days of landfall, his first trip to Puerto Rico is scheduled for next Tuesday, 13 days later.
Notice also that Trump found plenty of time to tweet about other topics in the period after Maria hit Puerto Rico, including 25 tweets about the NFL protests, 18 tweets about the UNGA, 15 about Senate health care bills and 13 about the Alabama U.S. Senate primary (including several he later deleted). During this same period he tweeted or retweeted a total 22 times about Puerto Rico and Maria (as of noon Sep 30), not including his five tweets this morning attacking the Mayor of San Juan and the media for its coverage of the crisis.
The data tells a story that is consistent with two especially troubling features of crisis management in the Trump Era: first, that Trump will only give a crisis attention once it starts to look especially bad on cable news (remember Syria?)—and, second, that this first feature means crises, even those affecting millions of Americans, may go ignored until its too late.
Zombie Trumpcare 2.0 has been laid to rest – again. Without delving into the policy or the politics of ACA repeal, this seems like a good time to touch on some notable aspects of the bill’s popularity.
It was striking how just how unpopular the bill was. The best illustration of this I have seen is from MIT political scientist Chris Warshaw, who compared numbers for the GOP health care reform to various major legislative proposals from the past few decades. GOP health care reform is less popular than them all – not just less popular than, say, the ACA (in 2009/10), but also less popular that TARP, which passed after a market crash and led to huge backlash, or Hillarycare, which failed to launch and helped foment the 1994 GOP wave.
Aggregating across polls, only 25% of public supports GOP health care bill. It remains historically unpopular compared to other major bills. pic.twitter.com/6rXPfbyguj
What’s at the heart of such historic unpopularity? Part of the blame surely falls at the feet of its low quality as legislation, a reality that’s manifested itself in the public eye through CBO scores, protests, industry opposition, and even skeptical Republican officials. But I wanted to focus on one Republican official in particular – the main one.
The reality is Donald Trump’s sales effort on this bill has been noticeably limp, not only when you compare it to Obama’s full-court press in 2009 and 2010, but also when you compare to his own remarks on things he actually cares about – like border walls, and stopping people from looking into his financial information. His remarks about health care have been rare, and when they appear they do nothing to directly address concerns about what’s in the new bill, but instead focus on Obamacare and Senate process.
I went to a recent YouGov poll to see if there was evidence of poor marketing. The basic idea is this: if Trump’s sales pitch on GOP health reform has been especially bad, it should mainly show up when looking at the voters who are in general likely to listen to Trump (e.g. Republicans, 2016 Trump voters). Conversely, on issues where he’s made a real push, you’d expect to see a boost with these same Trump-friendly voters.
A recent YouGov poll asked about a few Trump-backed policies, including the health care bill, the border wall, and the firing of James Comey. Below is public opinion on those three policies, broken down by party identification and 2016 vote. The policy changes are set alongside Trump’s own approval rating with the same groups.
To make it a bit clearer, here’s that data again, except this time it shows the difference between support for the policy and support for Donald Trump (negative numbers indicate the policy is less popular with a given subgroup than Trump is).
Support for the border wall and the firing of the FBI director track very closely to Trump’s own popularity among groups that mostly dislike him. These policy/personnel changes trail his own popularity only slightly among supporters.
The GOP health care bill is no different among Trump’s disapprovers – they like the bill as much (or as little) as they like Trump, no more, no less. Yet the the bill trails Trump by around 20 points when it comes to his supporters.
In other words, he appears to have effectively tied the border wall and the Comey firing to his own popularity, but has failed to do so with the Obamacare replacement. That’s still not necessarily a positive thing for the popularity of the firing or the border wall – Trump himself is not very popular, and there are compelling reasons to believe the border wall is actually less popular than it could be without its association to Trump. But a failure by Trump to even win over his supporters helps explain why the GOP health care bill is especially unpopular.
A British general election is tomorrow! That means lots of predictions and projections floating around for the next 24 hour. There are also some interesting questions surrounding British polling right now, after pollsters dramatically underestimated support for the Conservatives in 2015. But I have no prediction of my own; for one, I would find it hard to improve on what’s already out there. Here’s a smart piece on the state of things (along with an overview of changes to British polling post-2015) if that’s what you’re after.
I did find myself looking for a history of endorsements by UK newspapers, and couldn’t find anything recent. I have always found British newspaper culture fascinating, particularly for its open partisanship. While newspaper endorsements tend to reflect elite opinion in both the US and Britain, in Britain they are also much more likely to reflect an editorial line that has been visible in the paper’s tone and choice of coverage throughout the campaign. Probably the notorious endorsement to have taken place in modern British politics was the Sun’s in 1992, which led the tabloid to outright claim credit for an upset Tory victory: “it’s the Sun wot won it“. One study even found that the Sun’s shifting stances in 1997 (pro-Labour) and 2010 (pro-Conservative) were responsible for moving hundreds of thousands of votes.
So, where do UK newspapers find themselves now? After backing New Labour for several years, the Rupert Murdoch-owned Sun is back to supporting the Conservatives and had been reliably pro-Brexit. There are few other surprises compared to 2010, though it is interesting to see how some of the papers have evolved since the Labour landslide in 1997 (many through changing ownership, but also shifting tides).
The post-financial crash election of 2010 looks like a high water mark in newspaper support for the Conservatives since 1992. In fact, an analysis by LSE found that 71% of newspaper readership in 2010 went to a Tory-endorsing paper, more than double 2005. Of the 10 larger papers shown above, the 7 endorsing the Conservatives account for about 70% of the overall circulation (some of this could be overlap, and the list is incomplete, so classify this as very rough). Either way, May appears to have mostly kept hold of Tory gains under Cameron.
Of course, causation can go either way here – some readers may be swayed by the editorial line, and some newspapers may have made their picks to attract new readers or avoid alienating old ones.
Please let me know if you see any errors* or important omissions, some of this came secondhand. My main sources are here, here and here. In some cases they conflicted, and I did my best to find a better resource.
*The Evening Standard came out for the Conservatives today, so that’s been updated. It’s also been pointed out that the Sun endorsed SNP for Scotland in 2015.