Conor Lamb, the Democratic nominee for the March 13th special election in Pennsylvania's 18th Congressional District, talks about his campaign at his headquarters in Mount Lebanon, Pa., in February.
CNN  — 

Democrats are winning Republican-held seats in all sorts of elections since Donald Trump became president. They have flipped 39 state legislature seats. They won a US Senate seat in ruby Red Alabama back in December. They have not, however, actually won a seat in the US House that Republicans held at the beginning of Trump’s term.

That may change on March 13 when Democrat Conor Lamb takes on Republican Rick Saccone in the Pennsylvania 18th congressional district’s special election. The seat was vacated by Republican Tim Murphy following a sexual harassment scandal.

At first glance, the district looks bad for Lamb. Trump won it by about 20 percentage points in 2016.

Yet, most observers agree that even though Saccone is probably a slight favorite, Lamb has a decent shot of pulling off a stunning victory.

Here are five reasons to believe that Lamb may just win on March 13.

1. The polling is tight

A Monmouth University poll taken in the middle of February put Saccone up by a 49% to 46% margin. The average of all public polls taken in the race have shown a similarly tight race.

To put it mildly, a 3-percentage-point lead isn’t anything close to a sure thing in a House special election. I went back and looked at how off polls taken within three weeks of House special elections since 2004 were. Of the 75 polls in this database, the average absolute error was about 5 percentage points. The true margin of error (i.e. the 95 confidence interval) is closer to +/- 13 percentage points! You may remember when a special election poll taken with a few weeks of the Georgia 6th special election in 2017 had Democrat Jon Ossoff ahead by 7 points. He lost by 4.

The polling in Pennsylvania 18 indicates that anything from about a 13-point Saccone win to a 10-point Lamb win is realistically plausible.

2. The trend in special elections during Trump’s presidency

Democrats have so far been outperforming their presidential baseline by about 13 points in Trump-era special state and federal elections. If you were to take into account the Mitt Romney versus Barack Obama margin and the Trump versus Hillary Clinton margin in the districts and states with special elections so far and then tried to project how the Pennsylvania 18 special election would go, you’d expect Saccone to emerge victorious by less than 8 points.

But as I showed last week, Democratic candidates sometimes outperform and underperform this baseline projection. To boot, you’d expect Lamb to win about a third of the time given how other special elections have gone during the Trump presidency.

Further, Democrats have actually been doing even better than that 13 point overperformance looking at only the seven federal special elections held during the Trump presidency. Taking only those into account, Saccone would be favored to win by just over a percentage point. A forecast like that means this election is far too close to consider safe for Saccone.

3. Trump’s relative lack of popularity

Yes, Trump won by 20 points in Pennsylvania 18, but look at the Monmouth poll. His approval rating in the district is just 51% to a 47% disapproval rating. If voters were casting a ballot solely on how they felt about Trump, a tight race would be expected.

The 16-point drop in Trump’s net approval rating (approval rating - disapproval rating) in the district versus his margin in 2016 in the district fits with what we’re seeing nationally. In the latest CNN poll for example, Trump’s net approval rating with voters is -17 points. That is 15 points below his -2 point margin (i.e. loss in the popular vote) in the 2016 election.

All Lamb needs to take him over the line is a slight turnout advantage or for a few voters who like Trump casting their ballot for Lamb. Lamb is more popular than Saccone in the Monmouth poll, so this alone could put him over the top.

4. The Democratic Party registration edge

So far in special elections during the Trump era, it turns out that the 2012 presidential results in a district or state have actually been a far better indicator of the outcome than the 2016 results. That suggests that ancestrally Democratic areas that went heavily for Trump may be reverting back to form.

Pennsylvania 18 is an historically Democratic area. Democrat Walter Mondale won it in 1984, even as Republican Ronald Reagan was winning by nearly 20 points nationally.

Democrats still hold a 6-point party registration edge in the district, and the Monmouth poll projects an electorate that is 51% Democrat to 45% Republican.

Of course, Trump isn’t disliked in the district like he is nationally.

So to get an idea on how the party registration and presidential approval might interact with each other in Pennsylvania 18, I went back and looked at the last midterm election (i.e. when there was no presidential race on the ballot). Specifically, I examined the 2014 Cooperative Congressional Elections Study, a non-probability survey that gives a lot of pollsters heart palpitations. So, with that grain of salt, it asked voters (who registration records indicate voted) nationally their choice in that year’s House elections and their approval of the president. I only investigated those voters who had a party registration on file that was Democratic or Republican and were voting in districts in which a Democrat and a Republican were on the ballot.

To my surprise, party registration had a lot of explanatory power in a person’s House choice even after controlling for the voter’s feelings toward the President. Do keep in mind a special election can be a different beast than a midterm election. Still, the trends of 2014 indicate that a close race is to be expected in a district like Pennsylvania 18 where Trump’s net approval rating in the district is just above water and where Democrats hold a small registration advantage.

Keep in mind that in 2014, Murphy was unopposed. So this analysis doesn’t include him or Pennsylvania 18 in particular. When looking at only open seats, the analysis on the interaction between presidential approval and party registration holds.

5. The district is well educated

When some people think of southwest Pennsylvania (where Pennsylvania 18 is located) and the Rust Belt, they probably think of voters who lack a college degree. That’s part of the reason southwestern Pennsylvania went so heavily for Trump.

Pennsylvania 18 doesn’t really fit that mold, however. According to the 2016 American Community Survey, 37% of all adults 25 years and older in Pennsylvania 18 hold at least a college degree. That’s higher than any other congressional district in that part of Pennsylvania. It’s also higher than the 31% of Americans 25 years and older nationally who have at least a college degree.

That’s a good sign for Lamb in Pennsylvania 18 because Democrats are cleaning up among college educated voters nationally. The most recent CNN poll, for example, has Democrats ahead by 30 points among them on the generic congressional ballot. Even among white college graduates (who make up most of the college graduates in Pennsylvania 18), Democrats are leading by 19 percentage points.

In Monmouth’s poll, Lamb was up just 5 points among college educated voters. If he can push that margin just a little bit more towards the national margin Democrats enjoy among college educated voters, he’ll emerge victorious.