The predictive model I will use on Monday to get a feel for who will control the House is built around a boundary condition and proxy sampling.
The boundary condition, based on historical data, will set an upper limit for the best case Democrats can expect in the House.
I will use Senate races as a proxy sample for useful information about late trends affecting the House.
Senate races will be looked at from two perspectives to gather information about two categories of election trends affecting the House downstream from trends in the Senate –
1) Senate campaigns will be categorized by region of the country to get a feel for any national or regional trends breaking in favor of either Republicans or Democrats. From Senate trends I will estimate by proxy sampling trends for House races.
2) Senate campaigns will be categorized by whether the state holding an election is historically Republican leaning, Democrat leaning, or evenly divided. This category will be used to estimate by proxy levels of Republican and Democratic turnout in House races as well as estimate what direction Independents are breaking.
As for sampling methodology, if mainstream political analysts weren’t crackpots I would be surprised they adopted the crackpot idea that Senate trends are decoupled from House trends. Since they are crackpots I can only say they are staying true to form by wrongly thinking there is no statistical relationship between how contests will unfold in the Senate and House.
Pragmatically Distributed will assume there is a connection based on historical results. Past midterms show that outcomes in the Senate are mirrored in the House because competitive Senate races are geographically representative samples of the national vote.
The exception to this rule have been draws where there was no significant change in the partisan makeup of Congress.
This year’s Senate races are, as usual, geographically dispersed well enough across the country that I can use Senate trends in the last day or two as a proxy for the House. This is advantageous because direct polling in Congressional districts is tricky due to their often being drawn in odd twists and turns.
Let’s return to the boundary condition to create an estimate for a best case is for the Democrats. Based on that, what the Democrats need to win do is a heavy lift by historical standards.
The best point of comparison is the 2006 midterm.
In 2006, conditions were ideal for Democrats because the second Bush Presidency did everything it could to do what no Party should. In 2006 –
1) Democratic turnout was energized by the Iraq War.
2) Republican base turnout was depressed by Bush administration’s aggressive push to pass amnesty for illegal immigrants through Congress.
3) Independents went strongly against the GOP because of the unpopularity of the Iraq War and the Bush Administration’s botched attempt at Social Security reform.
The 2006 results resulted in Democrats winning a net of +31 seats and taking the popular vote for the House nationally by an 8 point spread, 52% to 44%.
To win the House this year the Democrats have to net +23 seats.
Remember they must net +23 seats, not win 23 seats. If Democrats take 27 Republican seats and Republicans take 5 Democratic seats the Democrats fail to take the House with a net gain of only +22.
Put another way, Democrats must do 74% as well as they did in 2006 under what were for them ideal electoral conditions.
Based on a statistical comparison of historical data (2006 compared to 2018), taking the House is an uphill battle for Democrats.
Although base turnout for Democrats will be high, Democrats cannot hope for depressed GOP turnout and cratering with Independents like they could in 2006.
Since Trump adopted my 3 proposals for the midterm campaign, Republicans have closed the gap in enthusiasm with the Republican base. Trump’s linking Democrat’s to their radical base has narrowed margins with Independents below the margins Democrats held in 2006.
I also suggested Trump gain with Independents by firing Mueller in order to trick Democrats into running on impeachment, which is electoral poison to Independents. He did not fire Mueller, but the Democrats did raise this same, electorally poisonous, idea by suggesting Brett Kavanaugh be impeached if they win the House.
The effect has been to energize GOP base voters who are already strongly approved of Trump’s policies and to alienate Independents from Democrats.
Using 2006 as an upper boundary, Democrats will not win get an 8 point spread they need to be confident of a House takeover.
A best case scenario for them would be a 6-7 point spread, which puts them just barely over the top or falling just short of House control.
If Democrats only win nationally by 5 points or less, Republicans will most likely hold the House.
To base my electoral predictions on I will be looking closely at whether the generic ballot narrows in favor of Republicans over the closing days, stays the same, or breaks in favor of Democrats.
Back to proxy sampling House races via Senate races.
My regional sampling of Senate races is as follows –
The Southwest – Arizona, Texas.
The Interior West – Nevada, Montana, North Dakota, Missouri.
The Southeast – Florida.
The Rust Belt – Indiana, Michigan, Wisconsin, Ohio, Pennsylvania, Minnesota.
The South – Tennessee.
The Northeast – New Jersey.
By covering a broad geographic area, any trendlines breaking in the Senate races should also spillover into downstream trendlines in the House.
I will be looking most closely (i.e., weighting) at Rust Belt trends since they host many key House races.
I am especially interested in using the geographic distribution to detect if a national wave is forming for Democrats, or Republicans, or no general trend is occuring.
If I see a trend breaking across multiple regions, I expect a wave building for either Democrats or Republicans.
If I see different regions breaking in different directions relative to others, then there is probably no wave for either party. No trend benefits Republicans since they win the House just by fighting Democrats to a stalemate.
The second category of Senate races, based on each state’s normal partisan leanings, will be used to estimate national House trends for Democrat and Republican turnout and any national trends, if any, among Independents.
Republican leaning states – Ohio, Texas, Tennessee, Arizona, Montana, North Dakota, Missouri, Indiana.
Evenly divided states – Florida, Nevada.
Democrat leaning states – Pennsylvania, Minnesota, Wisconsin, Michigan, New Jersey.
To indicate Republican turnout I want to see if Republicans how trending in solidly Republican states, Texas and Tennessee, Missouri, North Dakota, Indiana, Ohio.
If Republicans are secure in Republican leaning states, I will project Republican turnout will be good nationally for them House contests.
The same methodology will be used to project Democratic turnout in House races.
For Independents, I want to see whether there are any late breaking trends – or no trends – in swing states.
For Democratic states I only want to see if the Republican candidate is close, if the Dem is breaking away. In those Democrat friendly states, if I see Republican are closing (say, only down 5 points) that means Republicans are holding up well nationally with Independents and generating good GOP base turnout because the only way for a Republican to win those blue states is to win strong margins among Independents and get Republican voters out.
Pragmatically Distributed 