There is a particular satisfaction in watching a predictable system perform exactly as the literature said it would, though the satisfaction dims somewhat when one realizes the system in question is American electoral behavior and the performance involves the slow-motion disintegration of a governing coalition. The latest Economist/YouGov survey, dated August 13, places Democrats eight points ahead on the generic House ballot and President Trump’s approval rating at 33 percent. Neither figure is surprising. Both are, in the precise sense of the term, overdetermined.
Let us walk through the structural arithmetic, since the cable panelists whose segment brought this poll to wider attention, Anthony Fisher, Basil Smikle, Ahmed Baba, and Dan Nathan, sitting with Antonia Hylton on MSNBC’s Money, Power, Politics, are paid to react rather than to contextualize. The generic congressional ballot is a blunt instrument, but it is a blunt instrument with a respectable predictive record. In August 2006, Democrats led by roughly seven points. They picked up 31 seats. In August 2018, the lead was about four. They picked up 41. An eight-point margin in mid-August of a midterm year, with the president’s approval mired at 33 percent, is not a polling anomaly. It is a leading indicator of a tectonic event.
The presidential approval figure is the more structurally significant of the two, and not because 33 percent is a round number that looks alarming in a headline. The political science literature on midterm elections, from Tufte’s 1975 work onward, has established with considerable rigor that the single best predictor of midterm seat loss is the incumbent president’s approval rating. The mechanism is not mysterious. Presidential approval functions as a proxy for the public’s evaluation of the governing party’s stewardship. When that evaluation sours, co-partisan voters demobilize, opposition voters mobilize, and swing voters defect. The result is a transfer of seats that no amount of candidate quality or campaign spending fully counteracts. Trump at 33 percent is not a messaging problem. It is a physics problem.
What complicates the picture, and I do feel compelled to complicate it because that is what responsible analysis requires, is the structural distortion introduced by redistricting. Several discussants of this poll have noted that the “magic number” to overcome gerrymandering sits somewhere around six points. This is roughly correct, though the precise threshold varies by cycle and by state. An eight-point lead clears that bar, but not comfortably, and not in a way that any serious analyst would describe as safe. The Democratic Party’s own unpopularity, which is real and measurable, further muddies the waters. A wave election driven by presidential disapproval is not the same thing as an affirmative endorsement of the opposition. It is, rather, a negative verdict, and negative verdicts have a way of producing legislatures that lack the cohesion to govern effectively.
Shealeah Craighead (Public domain) via Wikimedia CommonsStill, the data are the data. YouGov, it should be noted, has been among the more conservative pollsters with respect to Democratic prospects this cycle, which makes an eight-point finding from that particular shop more rather than less credible. A pollster whose house effect leans Republican showing the GOP trailing by eight is the methodological equivalent of a defense witness who accidentally corroborates the prosecution’s timeline.
The deeper question, the one that the horse-race coverage will never ask because it lacks the conceptual apparatus, is whether the American electorate is punishing a specific president or rejecting a specific model of governance. Trump’s 33 percent approval is not merely a personal failure. It is the market clearing price of a political strategy built on grievance, institutional vandalism, and the systematic alienation of every voting bloc that is not white, male, and over fifty-five. That strategy delivered a coalition. It has now consumed it. The midterm results will be the receipt.