Eh, There have been maybe less than a half dozen times in the last century where the party in power in the White House didn’t lose major seats in a mid-term. And none had a president this unpopular.
I think 1934, after 9/11, 98 with Clinton and maybe 1 or 2 more.
Just to note on this, the big "losses" historically (45+ House seats) in modern mid-terms since Gallup started are (all Gallup figures to be consistent):
1938: FDR hit his all time greatest disapproval of 46% three times in this year including the week of the election
1942: The real outlier historically, FDR had an approval rating of 70% at the time, Republicans actually won 51% of the popular vote, and I think this had more to do with 1940 blocking a continued "correction" of the original FDR landslides of 1932/1936...i.e. there was no attrition of Democratic over-representation in seats, so they blocked up for another election (also, anti-war voters?)
1946: Truman fell from 50% to about 35% approval by the start of the year to the election...after the election he actually rebounded to 60% and the highest point he would hold post-war except for a similar post-re-election spike
1958: The only year Eisenhower spent under 60% approval, including earlier in the year hitting his lowest point of 47%
1966: Like Ike, the first year LBJ spent under 60%, he was under 50% for the entire second half of the year
1974: Doubtfully related to Ford's approval (55% on election week for example), probably had more to do with that other guy...
1994: Clinton had fallen from 60% to start the year to hit 40% in September, he rebounded a tiny bit in October, just to hit 40% again after the election the lowest point he would ever be for the rest of his term...Clinton's actual lowest point was in summer of 1993 when he spent a month under 40%
2010: Obama had fallen to his lowest point of 45% before a slight rebound and then re-drop similar to Clinton had in 1994...Obama's worst stretch was late 2013-2014, spending much of it around 40%, which didn't lead to a House landslide as GOP had maxed out nearly, but is a backwards explanation for how the GOP ran the table on the Governors and Senators much to the suicidal exasperation of PoliGAF on election night
2002 is funny because technically W. had the largest start of the year to election drop ever for a mid-term, but he started at almost 95% so it like totally doesn't count. He wouldn't fall under 40% until late 2005.
Trump presents an interesting thorn in the side of the midterm theory because of one simple fact. His approval has been so low from the start he has borderline nowhere to drop. His HIGH in the RCP average is 46%, his LOW is 37%.
spoiler (click to show/hide)
This is doubly interesting because of how it relates to Obama's approval history, counting from when Obama first went negative in the RCP average and then back to positive (i.e. when he stopped dropping from his electoral high, much like I would count from when W. finally stopped dropping from 9/11 temporarily) up to 2016 (when politics turned fully to his successors) his approval LOW was 40% and his approval HIGH was 54% on Christmas post-re-election. If we toss out the "era of good feelings" of December 2012 and January 2013 and the week bin Laden got got, his HIGH was 50%.
Looking at RCP's data for the last three Presidents there's arguably really only a maximum ten point swing in any given year between a President's HIGH and LOW points. Especially if you account for outlier events. As I'm starting to ramble in my nonsense numbers way I'll put this under spoiler tags. But looking at our three Presidents this century, their year average, their high and low points in that year, in approval...
YEAR: AVG | HIGH | LOW
2001: 66 | 60* | 52* (actual HIGH of 89 on start of Afghan invasion but I didn't count anything after 9/11)
2002: 71 | 82 | 62
2003: 59 | 72* | 51 (*Iraq invasion, otherwise 65)
2004: 50 | 53 | 45
2005: 45 | 51 | 40
2006: 39 | 43 | 34
2007: 34 | 37 | 30
2008: 30 | 36 | 25 (it actually went up for the economic collapse, and then stopped dropping after Obama elected)
2009: 57 | 65 | 49
2010: 47 | 50 | 44
2011: 47 | 53 | 42 (only 53 for bin Laden!)
2012: 49 | 54 | 45 (54 for getting re-elected! 50 was his high for the year after the second debate)
2013: 46 | 53 | 40
2014: 43 | 44 | 41
2015: 45 | 46 | 43
2016: 50 | 57 | 45 (high of 51 before Trump elected)
2017: 40* | 46 | 37 (i rounded down, and then subtracted one...for reviewers tilt...or to toss out his "era of good feelings" first month and a half)
tl;dr i have no idea what this means for midterms...er wait that's now how pundits work, then again, this is the era where Bill Kristol called an election and got it nearly right margin wise
p.s. the trend actually continues farther back than W. which leads me to a more sensible conclusion, more polling of which is constant with numerous tracking polls, stamps out the swings and renders the topline data as nearly as useless for election predicting as Congress' approval rating condescending question mark
