Trump isn't everything in the midterms. Local factors will matter too

(CNN)Former Speaker of the House Tip O'Neill made the phrase "all politics is local" famous. The implication being House Democratic or Republicans members could win in districts in which their party's presidential candidate did poorly.

O'Neill's wisdom, however, seems to have been pushed aside in recent years. The presidential voting pattern of a district has become a very strong predictor of how it will vote in House elections. That's why you'll consistently hear folks like myself cite the presidential voting patterns and pretty much only the presidential voting patterns of a district to help explain why it votes the way it does.
The only problem with what analysts are doing is that it's probably wrong, as FiveThirtyEight's Nate Silver pointed out last week. Although politics have become more presidential, local voting patterns still give us additional information about a district's partisanship.
Add CNN's 2018 elections calendar

Stay up to date on key election dates and other events, such as town halls and debates, as they get added throughout the year. Subscribe on your Google or Apple calendar.

When you look at the race raters' list of competitive races in 2018, you see a lot of districts where Hillary Clinton did well and where Republicans are in trouble. You also see districts such as California's 39th and 49th, where Barack Obama lost and Hillary Clinton easily won.
    Some ratings, though, look out of place. President Donald Trump won districts like Kansas 2 (lean Republican in CNN's ratings), Kentucky 6 (lean Republican), North Carolina 9 (toss-up) and West Virginia 3 (toss-up) by 18, 15, 12 and 49 points, respectively. These districts also voted for Mitt Romney by double digits, so these are solid Republican districts on the presidential level.
    But now look at how Democrats have been doing in local and state races in the four districts. The Democratic firm TargetSmart provided me with a measurement that is essentially the average local and state Democratic performance in these districts over the past few cycles. In all of them, the Democrats outperformed the 2016 presidential baseline by at least 5 points. In all but North Carolina 9, the average Democrat outpaced Hillary Clinton by at least 10 points. In West Virginia 3, the average local and state Democrat lost this district by less than 2 points even though Trump won it by 49.
    In other words, these are all places where local and state election results suggest that given a strong Democratic national environment, Democrats should have a real shot of winning a House seat in 2018 even if the presidential baseline is less optimistic.
    Fortunately for Republicans, this goes the other way too. There are some Republican-held districts where members are hanging on when perhaps you would expect otherwise. Race raters have Florida 26 as at worse a toss-up for Republicans and New York 24 as likely Republican. That's not what you would think given that Trump lost both of these districts by greater than he lost nationally. In Florida 26, he was defeated by double digits. Romney lost both districts by double digits.
    Again though, look at the local elections. The average Republican in local and state elections lost in Florida 26 by only 2 points. The average Republican in local and statewide elections actually won in New York 24 by 2.
    Now, it's obviously no guarantee that Republicans will hold on in these districts. Just like it's not guaranteed that Democrats will win or come close in the districts above. The presidential vote may end up totally trumping local voting patterns.
    That's not what happened in 2016, however. I went back and looked at the margin in the House races where one Democrat and one Republican ran (outside of Louisiana, where I have no state voting data). No matter how you measured the presidential voting patterns of a district (just looking at 2016, averaging 2012 and 2016, a weighted average of 2012 and 2016 where 2016 was given more weight), state elections still held some explanatory power.
    For example, to best explain the House vote in 2016 using just the 2016 presidential results and the local and state election results over the past few years, you'd weigh the 2016 presidential results at only 1.5 times the local and state results. That is, you want to look at both presidential and local and state results. This, of course, is in the average district. In some districts like Florida 25, the 2016 presidential result was a terrible baseline, and state and local election results were more predictive. In others like Oklahoma 2, the presidential vote was far more explanatory.
      It's not always clear why the presidential vote is more important in some districts and less important than others. The quality of the candidates on each side of the aisle certainly plays some role. Candidates with greater ties to the local community (like Conor Lamb earlier this year) might help localize a race, for example.
      What is clear is that looking only at the past presidential vote is a mistake for knowing whether or not a district will end up being close. That's probably good news for Democrats. It means a wider playing field for them, especially given that more congressional districts leaned Trump in 2016 than the national vote implied.