Popular punditry predicts a Democratic wave in the upcoming midterms. As I usually have, I will be betting against them. Here are a few reasons why.
First, historically, wave elections of the size and scale needed for the Democrats to retake Congress are exceedingly and exceptionally rare. Second, the measurements relied upon by the political punditry are notoriously inaccurate in forecasting the actual result at the real polls. Third, the absence of a recession, resignation, war or realignment, make the probability of a wave very low. Fourth, and finally, the geography and demography of the parties’ constituencies disfavors Democrats in midterm elections.
Remember this above all: the Democrats need to improve on their last midterm popular vote margin by 13 to 14 points, and such a massive shift has only occurred when a realignment of a large group of voters occurred. The “enthusiasm gap” matters, but it averages to about a 5 point swing, nowhere near enough to give the Democrats Congress in 2018.
What is the norm? 2018 represents a change of Presidential party power, an event that occurred 9 times since 1934. Measuring midterm swings is best done by comparing midterms to midterms, not midterms to presidential-year elections. By that measurement, here has been the party swing for those 8 elections:
Change In Midterm Presidential Party Popular Vote Margin From Prior Midterm After Change in Party in White House
- 1954: -5
- 1962: -7
- 1970: -6
- 1978: +0
- 1982: -3
- 1994: -15
- 2002: +4
- 2010: -15
The norm is only a swing of about a 5 point loss for the Presidential party. This is the popularly discussed “enthusiasm gap” which is real, and recurs in most midterm elections. The anger of the party-out-of-presidential-power tends to trump the satisfied mindset of the party-in-presidential-power voters in terms of turnout. That said, that kind of swing —5 to 6 points in the margin — is too small to create the kind of wave Democrats need in 2018.
What would a “typical” midterm drop-off mean for the Congressional GOP in 2018? A popular vote in the House about the same as 2016, with few GOP seat losses, as the 5 point decline in 2018 has already been “baked in” from 2016, due to the presidential year voter turnout groups favoring Democrats in presidential election years. Indeed, big swings between midterms rarely occur in general. Here are the numbers:
Change In Midterm Presidential Party Popular Vote Margin From Prior Midterm
- 1942: -5
- 1946: -5
- 1950: +10
- 1954: -5
- 1958: -6
- 1962: -7
- 1966: -2
- 1970: -6
- 1974: -8
- 1978: -8
- 1982: -3
- 1986: +2
- 1990: +2
- 1994: -15
- 1998: +6
- 2002: +4
- 2006: -13
- 2010: -15
- 2014: +2
Note: a change of 14 points or more between midterms has only occurred 2 times in 19 midterms, and in both cases major realignments occurred: in 1994, evangelicals left the Democratic Congressional party en masse; and in 2010, Jacksonian America in Appalachia and the upper south left the Democratic Congressional party en masse.
Without a realignment, only two elections showed double-digit change between midterms: one with a deeply unpopular president gaining votes (1950) and the other with a deeply unpopular president in the middle of an unpopular war losing votes (2006). The controlling factor appears to be war, not presidential job approval, for those rare non-realignment double-digit midterm swings.
Second, many pundits rely upon dubious measurements for forecasting midterm elections: generic ballot polls, presidential-year voting patterns, presidential job approval. None of those proves reliable, especially in the modern era of polling. (With the notable exception of PPD, few pollsters proved accurate in the the low-response-rate, cell-heavy era of post 2013 polling.)
The generic ballot is mostly useless, as Roper’s generic ballot reveals. Here is the average overstatement of Democratic support in Roper’s generic ballot before the fall of the election year:
Dem Overstatement in Roper Generic Ballot Polls
- 1966: +12
- 1970: +7
- 1974: +15
- 1978: +11
- 1982: +4
- 1986: +0
- 1990: -2
- 1994: +11
- 1998: +6
- 2002: +3
- 2006: +4
- 2010: +13
- 2014: +10
Comparable problems infect media polls. ABC forecast big Democratic generic ballot wins in 1994, 2010 and 2014, all years the GOP swept with safe margins. CBS’ 2014 polls showed Democrats with a consistent lead in the generic ballot, only to lose by 5 points. In the 100+ media polls since 1994, the average media poll overstates the Democratic generic ballot success by at least 7 points.
This only gets worse in an era of unrepresentative media polls that overstate Democratic voters due to the low response rate to modern phone polls, the over-reliance on cell phones, and the inadequate weighting and post-stratification methods utilized by many media pollsters. Just take a glance back at the 2014 Senate polls, and you will see how badly unreliable such media-friendly polls have become.
Presidential job approval proves equally dubious as a measurement.
Truman’s job approval mired in the 30’s before the 1950 midterms; his party did 10 points better than the prior midterm. Democrats saw no drop-off from the pre-Watergate midterms under Nixon despite Carter’s sub-50% job approval. Democrats dropped a mere 2 points off of their 1962 Congressional support despite the massive difference in job approval between Kennedy and LBJ. Republicans experienced little drop-off in 1982, even with Reagan’s frequently 30s-level job approval ratings.
Equally, Eisenhower’s high ratings couldn’t stave off back-to-back drop offs in his two midterms, nor could Nixon’s good 1970 numbers stave off losses, just as Clinton’s high 1998 ratings failed to restore the pre-Clinton Democratic congressional voting edge. Unpopular war (2006), resignations (1974), and recessions (1958) play a key role, and occasionally, pundits confuse job approval ratings with those events; job approval, by itself, has little correlation with midterm election outcomes.
A few examples:
Presidential Party Midterm Drop-off in Popular Vote
- 1950 (better than 1942, despite low Truman ratings)
- 1958 (double-digit drop off from 1950, despite high Ike ratings)
- 1966 (little change from 1962, despite low LBJ ratings)
- 1978 (no change from 1970, despite low Carter ratings)
- 1982 (modest change from 1978, despite low Reagan ratings)
- 1998 (9 point drop off from 1990, despite high Clinton ratings)
The one place presidential job approval appears to provide some limited value is job approval on the economy and security. Notably, in both categories, Trump continues to way over-perform his job approval numbers on issues of economy and security, where he consistently enjoys a favorable and positive rating from even Democratic-inflated media polls.
Third, nothing suggests a major demographic shift between the parties in Congressional voting patterns. In true realignment elections, a significant group of voters shift their partisan preference. This is usually forecast by changing trends in partisan voter registration and usually captured by Gallup’s party identification polling.
In this case, the voter registration trends actually favor Republicans in almost every swing state, and most states of the country, while Gallup shows no real change in party self-identification amongst the public since 2016. Pundits forget that Democratic Congressional majorities always depended on the deep Congressional voting loyalty of the oldest group of Democrats in America: what Sean Trende calls “Jacksonian Americans.”
This group of working class voters from the upper south, Appalachia, and their culturally kindred, formed a geographical reverse L, from the plains of eastern Oklahoma through the countryside of Arkansas, from western Tennessee to western North Carolina, dipping into the piney woods and upcountry of Mississippi, Alabama, Georgia and South Carolina, dancing along the gulf coasts of cajun Louisiana to the redneck riviera in the Florida panhandle, up through the spine of coal country in eastern Kentucky, West Virginia, and the mining-influenced Scotch-Irish, Welsh, old ethnic towns of southern Ohio and large swaths of Pennsylvania.
Before the 2010 elections, Democrats held nearly 90% of the Congressional seats in that area; after 2010, and ever since, they have been lucky to get more than 10% of the Congressional seats there. Note: the Democrats have never held a Congressional majority without those seats. Ever.
Fourth, and finally, the identity politics of the Democratic party and self-segregation of their voting constituencies isolates them in a state-driven Congressional voting process. While Trump voters can travel coast to coast, border to border, without ever crossing a Clinton county, Clinton voters need private jets to hop and skip across the isolated archipelago of their own voters.
This creates a geographic edge for the Republican Congressional majorities, forcing Democrats to likely need an 8 to 9 point national victory in the popular vote to even have a shot at majority control of either branch of Congress.
Then, there’s the demographic problem.
From the 1930s until 2010, the Democrats enjoyed a demographic edge in midterm elections due to one key factor: age. The older a voter, the more likely they were a Democratic loyalist. No more. Now, as you scale the age spectrum, the more Republican the voter group. Democrats depend upon historically high turnout of historically low turnout groups — millennial and minorities — to win back Congressional control, a simply improbable outcome from a historical perspective.
Without war, recession, resignation or realignment, the pollsters, pundits, and professional predictors look likely to get it wrong again in 2018. Don’t bet on a Democratic Congress in 2018. At 3-2 odds, my money is on the other side.
Robert Barnes is a high-profile criminal defense lawyer and has been dubbed by the media as America’s most successful political gambler. Learn more at Barnes Law LLP.