The race was the latest example of increased Democratic energy and tepid Republican candidate performance that has defined recent special elections, including House races in Arizona and Pennsylvania and a Senate race in Alabama.

Despite Republicans dramatically outspending Democrats and a last-minute visit by Trump, Republican state Sen. Troy Balderson received just 1,754 more votes than Democrat Danny O’Connor, an elected county recorder, with thousands of provisional votes still outstanding.

The margin of less than 1 percent came in a district that Trump won by 11 points in 2016 and where Republicans have held control since 1983. If the vote difference tightens further, it could trigger an automatic recount under Ohio election law.

In July of 2006, The Cook Political Report rated just 14 GOP-held seats as highly vulnerable. By November, the number of GOP-held seats in danger had tripled to 43. We saw a similar pattern in 2010. In August of that year, we listed 36 Democratic-held seats as highly vulnerable. By November, the number of vulnerable Democratic-held seats had more than doubled to 78. On Election Day of 2006, Republicans lost 30 seats; Democrats lost 63 seats in 2010.

(snip)

There wasn’t a point where the bottom just dropped out for one party. The approval rating for President George W. Bush was 40 percent in mid-July of 2006 and 38 percent in early November. President Obama was sitting at 44 percent in mid-August of 2010 and 45 percent in early November.

Nor was it that a bunch of ‘new’ districts came into play between the summer and fall of those years. By the summer of 2006 and 2010, the Cook Political Report had already identified as shaky or potentially weak all but a handful of the seats that ultimately fell that fall.

Instead, as summer turned to fall it became more and more clear that things weren’t going to get any better for the party in the White House. And, as such, races that we listed as potentially vulnerable in the summer were moved into the more highly vulnerable category by fall.