Ronald Brownstein's new piece on the CNN site has a very good rundown of the very good and not so good signs for the Democrats moving into the 2018 election.
The two big strengths (very interesting data here; note the stuff on white college men and, especially, white noncollege women):
"1. The white-collar suburban discontent with Trump is real and widespread.
The shift away from the GOP among white voters holding at least a four-year college degree is most intense among women, but also apparent among men. And those voters are retreating from the GOP not only along the East Coast (across Republican-held suburban seats in New Jersey, Pennsylvania and Northern Virginia) and the West Coast (in a concentration of five GOP-held seats around Los Angeles and another near Seattle) but importantly also through the center of the country. There, Democrats are poised to capture suburban seats outside Minneapolis, Kansas City, Denver, Detroit, Chicago and Tucson; have toss up chances in other seats near Des Moines, Salt Lake City, Detroit and Chicago; and have solid, though more challenging opportunities in Houston and Dallas. (More on that below.)
When the Washington Post/Schar School poll recently surveyed voters in 69 of the most competitive House districts they found that Democrats led among college-educated whites in them by fully 13 percentage points; by comparison, House Republicans carried those voters by nearly 20 point margins in both the 2010 and 2014 mid-terms, according to exit polls. Republicans respectively won control of the House and Senate in those midterms.
2. Democratic Senate and governor candidates in the Midwest are showing renewed competitiveness among blue-collar white voters who keyed Trump's victories in the states that propelled him into the White House.
Democratic Senate incumbents in Ohio, Wisconsin, Pennsylvania and Michigan -- all states won by Trump -- now appear solid favorites for re-election. The party is favored for the governorships in Michigan and Pennsylvania and locked in close races in Wisconsin, Ohio and Iowa -- the fifth Midwestern state key to Trump's 2016 victory. And it could pick up as many as four House seats combined in Iowa and Michigan.
In each case, that's at least partly because the Democratic nominees are posting much better numbers than Hillary Clinton among working-class white voters. Some of that may reflect what political professionals call "differential turnout" -- meaning that the non-college whites who dislike Trump are more likely to show up than the working-class whites who surged to the polls for him in 2016, but aren't as enthusiastic about conventional Republican candidates.
But Trump also appears to have suffered genuine erosion among working-class white women, largely because of his attempt to repeal the Affordable Care Act, and a sense among many that the improved national economy hasn't provided them appreciably more security. If that crack in Trump's armor persists to 2020, it would arguably provide the single most important advance for Democrats in the midterm election."
And the big weakness:
"1. Trump 's provocations alone show few signs of improving the subpar turnout patterns among Latinos and millennials, two core Democratic constituencies.
In polls, both groups express preponderant opposition to Trump's posture on cultural and racial issues. But most polls suggest that their turnout next month will plummet compared to 2016, just as it typically has in midterm elections. Compounding the problem, when Latino turnout sags, what's left in the voter pool tends to be older and more Republican.
Democrats received encouraging news from Sunday's ABC/Washington Post poll, which found much higher levels of youth engagement than almost any other recent survey. But that result looks like an outlier compared to most other polls. And even if young people participate in somewhat higher numbers, their share of the vote could fall if they don't keep pace with the greater-than-usual midterm interest evident among other voter groups. By 2020, millennials will significantly exceed baby boomers as a share of eligible voters, but based on their turnout trajectory they will continue to lag them among actual voters. That would be a huge opportunity cost for Democrats given Trump's consistently low marks with the generation (apart from younger non-college whites)."
A caveat on the youth turnout observation. Geoffrey Skelley on 538 has an interesting piece where he makes the case that youth turnout, based on some other indicators, might actually be pretty good this year. He says:
"Looking at the historical trends, there’s no question that youth voter turnout is consistently low in midterms, but exit poll data from competitive statewide elections in 2017 suggests that 2018 could set a record high for young voter participation....Polling from the Harvard Kennedy School’s Institute of Politics also gives us reason to believe we may see high turnout from young voters. The institute conducts a long-running, large-sample poll of young Americans...[I]n the IOP’s spring 2018 poll, 37 percent of 18- to 29-year-olds answered they would “definitely” vote, which was a new record high."
I don't know if this is right or not, since the data he cites are hardly definitive. But it's definitely interesting and suggests Democrats should not give up hope for decent youth turnout this cycle.