Help me keep The Crosstab ad free. Subscribe to premium content and join the community on Patreon.
Become a Patron!

What the Pennsylvania Congressional Map Redraw Means for 2018

What the Pennsylvania Congressional Map Redraw Means for 2018

After Pennsylvania Democrats and Republicans sparred over the court-mandated redraw of the state’s congressional district map, the Pennsylvania Supreme Court finally released their remedy. The new map is a full-frontal rebuttal of Republican partisan gerrymandering in the state and lends a helping hand to Democrats who hope to build their 2019 U.S. House majority partly off of the suburbs of Philadelphia and seaside. If the new map were in play the last election, Hillary Clinton would have won more votes than Donald Trump in eight of the state’s eighteen districts (rather than the six she did win) in a statewide contest in which President Trump won by just 0.72 percentage points. The new map is a boon to Democratic prospects in November.

This post elaborate on tweets I wrote yesterday about the new PA map. You can view (and retweet) those at this link:

What Changed?

So what exactly does the new map change? In short, it’s a complete redesign of the state’s congressional district plan — complete with redrawn renumbered districts — that spreads Democrats out more equitably in suburban districts around Philadelphia and Pittsburgh. The end result is a plan that increases their share of the vote in 4 competitive districts (flipping two from red to blue), places more Republican voters in some GOP strongholds in western and central Pennsylvania, and create two additional competitive districts in Alleghany and Dauphin Counties. The map below was made by Decision Desk HQ contributor J. Miles Coleman and shows these differences beautifully:

Above, we see the changes in district-drawing are mostly subtle, but they combine to make a big impact. In the figure below I plot each district to show its new vs old vote margins for Hillary Clinton in the 2016 U.S. presidential election. The higher a point falls above the diagonal black line, the larger increase in Clinton vote margin between the old and new map.

Take note of the points that fall directly below the horizontal, dashed gray line (around -20 on the x-axis). These two districts (numbered 11 and 16 on the new map, 4 and 3 on the old map) provide solid opportunities for Democrats to pick up seats in November, and they account for one additional competitive seat that was not present in the old map. This will be a very big deal when I get to talking about simulated outcomes and probabilistic expectations in the latter half of the next section, so remember this.

For now, let us turn our attention to the 2018 elections. What do an additional two seats mean for Democrats?

What Does That Mean for 2018?

A two-seat Democratic pickup is, frankly, not that large in the grand scheme of things. There are 435 House seats, after all, and Democrats need 24 to take over the House majority (today, my model thinks they’ll win 27). But 2 seats could very well make the difference this November: in a tied race, a buffer zone in Pennsylvania could provide Democrats some extra wiggle room with the pickups they’re hoping to score in states like Florida, Texas, and California.

To be sure, two seats won’t move a tied race from, say, 50% to 70% chance of a Democratic majority in the U.S. House — but it could move things from a 55% race to a 60% race. And five percent isn’t nothing; 5% made up about one-third of the chance that Donald Trump had of winning the 2016 presidential election.

Before we visit the race in aggregate, I think it’s fitting to drill a little bit lower into the probabilities that lie within the new PA map.

Taking a similar approach to my post on the GOP-proposed PA congressional map remedy, I can figure out what the new map means for the expected number of seats Democrats may win (and in different national environments) in Pennsylvania. I use an abstracted version of my full 2018 forecasting model to predict only the seats in Pennsylvania — using the new 2016 Clinton data and incumbency status for each district — and then simulate the election thousands of times to produce a distribution of possible PA seats. I repeat this process for a host of different national environments: one group of simulations each for a hypothetical world in which Democrats win by 0, 1, 2, 3… or 15% in the national popular vote. What we learn is striking:

The plan is thus a huge win for Pennsylvania Democrats, providing them with an upshot even beyond the two-seat gain they are likely to see outright. This is, again, almost entirely due to the addition of more competitive districts in Pennsylvania. And the additional seats for Democrats -— 2.6 seats on average in the simulations, if we’re being precise — goes directly into their national expectation as well.

But what about national Democrats? This conversation about state-level probabilities is not exactly helpful. I want hard numbers! Facts! Real data! So I turned to my forecasting model — with this distributional shift in expected post-2018 Democratic seats in hand — to determine the impact under different national environments. (Note: so far, I’ve been talking about “national environment” in terms of actual votes cast. Now, because we’re forecasting, I use “national environment” to mean the generic ballot polling average on election day.)

In effect, I matched the current seats-votes-probability curve with a revised one for the new maps, which I obtained by re-simulating the national U.S. House midterm hundreds of thousands of times, under different swings based on national polls, with the new district-level estimates for partisanship and incumbency status.

Simulated 2018 House Probabilities Under Different PA Congressional Maps
Chance of Dem. House Majority
Dem. House Margin (%) Old Chance New Chance Difference
0 20 22 2
1 23 25 2
2 26 29 3
3 30 33 3
4 34 37 3
5 38 41 3
6 43 46 3
7 47 51 4
8 51 56 5
9 57 61 4
10 60 65 5
11 66 70 4
12 70 73 3
13 74 77 3
14 78 80 2
15 81 83 2

These new simulations reveal two important considerations for benchmarking 2018. First, the 2-3 seat median gain from the Pennsylvania redraw translates to a 3% average increase in win probability for Democrats. Second, that increase in win probability is actually higher, roughly 4-5%, if generic congressional ballot polls point toward a 7-8 point race. This makes sense; as I said early, an extra two seats would matter quite a bit for the Democratic Party if that’s all they need to win back the majority, but 2 seats offer diminishing returns if they’re already up (or down) by 50-60 seats. In summary, the closer we are to a tied race, the more the new PA congressional boundaries matters.

On top of these probabilistic concerns, Democrats will be happy to know that the updated models say they’ll be favored to win with just a 6.8% vote margin in national polls, instead of the 7.7% necessary with the old Pennsylvania map.

But two seats, in the end, is just two seats. There are a host of other moving parts in a U.S. House election, including 433 other Congressional districts, error in national vote estimation, surprise candidates that won’t be viable. If Pennsylvania’s two seats do (or don’t) swing to the Democrats, they could be easily canceled out by an MN-01 or NV-03 that doesn’t (or does).

It is hard to strike the right balance between blowing the Pennsylvania map out of proportion, based on state-level analysis, and minimizing its changes, based on national contextualization. The consequences of the revision are undoubtedly significant; it shifts expected districts 2-3 seats (and expected win probability 3-4%) toward Democrats, but it’s likely not going to change how Democrats are playing the 2018 midterm game. They still need a goalkeeper, — perhaps 10 seats total in Texas, California, Arizona, and Florida — some defense, — pickups in 5 or so New Jersey and Midwestern districts — offensive players, — a handful of white districts in Colorado, Washington, and New York — on top of their striker — Pennsylvania districts. And (if you’ll let me assault this metaphor even more) a team rarely plays well if the crowd isn’t cheering it on. And this year, the other team is not likely to give them any penalty kicks.

The Pennsylvania map is a big gain for Democrats that comes at a time when Democrats need a big gain, but the overall dynamics of the 2018 race remain consistent. And, of course, it’s still a long 269 days until November 6, 2018. Until then, make sure you follow along with the 2018 forecast model as it continues to labor on.

G. Elliott Morris avatar
About G. Elliott Morris
Elliott is a ("big") data journalist and data scientist who specializes in American politics, public opinion, and predictive analytics. He is a government, history, and computer science student at The University of Texas at Austin and worked previously for the Pew Research Center and Decision Desk HQ.
comments powered by Disqus