2018 Midterm Election Prediction as of Sept 20

I made a simplistic model that looks at generic ballot polling versus number of seats. My forecast based on a normal election year is Democrats pick up between 14 and 17 seats. If the Democrats enjoy a wave similar to Republicans in 2010, the may pick up as many as 50 seats. If they enjoy a wave as in 2006, 21 seats. The average of the past four midterms, which includes the GOP outlier blowout of 2010, projects 23 seats. Democrats need 23 to take the House.

Individual House races can swing on the quality of the candidates. Turnout plays a huge role. Democrats have a demographic advantage in states such as Texas and Arizona, states that will likely turn blue over the next decade. Big turnout could make that shift come earlier.

I'm a little surprised at the projection because it says the House is close to a tossup today. Republicans typically surge late in polling though (and they usually outperform 2 to 3 percent on Election Day). FiveThirtyEight forecasts 80 percent odds of Dem takeover with an average gain of 37 seats.

If the Dems widen their polling lead in the next few weeks, the odds of a House takeover will be solidly in their favor. If the polling tightens, a GOP hold starts looking possible.