Does gerrymandering of some US electoral boundaries drive polarisation between parties?

Question

Political control of redistricting in state and congressional elections is common in the US. I've already addressed one possible effect in this question: Does politically controlled redistricting in US elections make incumbents more secure?

One of the possible side-effects of increased security is that incumbent politicians don't have to appeal to the "floating voter" in the middle of the political spectrum. Some commentators speculate that this has driven increased political polarisation between the main parties. For example, this Huffington Post article about increasing political polarisation argues that incumbent security is one significant cause:

About 80% of all Congressional Districts are solidly Democratic or Republican. We have bipartisan agreement on one issue -- politicians prefer safe Congressional seats, which sometimes happens naturally, but also because politicians use redistricting to create them. Instead of engaging with challengers from the other party, most members of Congress now engage only with challengers within their own party -- Democrats with their Left, and Republicans with their Right. Bipartisan compromise becomes a dirty word, as a consequence.

It is clear that polarisation is increasing. See this chart for a recent summary (sourced from thedailyviz):

However, though it seem obvious to many, not everyone agrees that it drives polarisation, for example:

Both pundits and scholars have blamed increasing levels of partisan conflict and polarization in Congress on the effects of partisan gerrymandering. We assess whether there is a strong causal relationship between congressional districting and polarization. We find very little evidence for such a link.

So what does the balance of evidence look like? Is there a relationship between higher security in gerrymandered constituencies and the polarisation of political views?

Answer 1

Here is a Nate Silver article on the subject. He talks about the dwindling number of swing districts. When congressman are in landslide districts, they don't face an electoral threat in general elections, but still have to worry about primary elections. Some of the decline in the number of swing districts can be attributed to redistricting, according to the article:

Some of this was because of the redistricting that took place after the 2010 elections. Republicans were in charge of the redistricting process in many states, and they made efforts to shore up their incumbents, while packing Democrats into a few overwhelmingly Democratic districts.

However, gerrymandering is insufficient to explain all of the increased partisanship in Congress. Increased polarization exceeded the polarization caused just by gerrymandering:

However, a more careful look at the chart reveals increasing polarization. The slope of the black regression line in the chart is greater than one (specifically, it is about 1.08). In plain English, this means that polarization increased by about 8 percent from 2008 to 2012 — above and beyond any changes brought about by redistricting.

This doesn't mean that gerrymandering doesn't affect elections. Gerrymandering helped Republicans win a majority in the House of Representatives in 2012, even though Democrats won more votes. However, the effects of gerrymandering are insufficient to fully explain the increase in partisanship.

Answer 2

No. This has the causality backwards. Increased polarization leads to unbalanced vote shares. See for example, the Real Clear Politics published Gerrymandering Isn't to Blame for D.C. Impasse:

Again, the point here isn’t that gerrymandering hasn’t had any effect on party polarization. It is just that the effects are likely very small. What’s really happened, more than anything else, is that conservative areas of the country have, at least for now, become extremely reluctant to elect conservative or moderate Democrats, while liberal areas have largely given up on liberal or moderate Republicans. This has resulted in party caucuses that are increasingly made up of ideologues, and has made political compromise difficult. If there’s anyone to point the finger at, it’s ourselves.

Author Sean Trende identifies four reasons to think that gerrymandering isn't the primary driver:

1. The disparity between votes and seats is an endemic feature of U.S. politics.

Basically, this has been true for a long time. It is actually less true now than it was in less partisan times, e.g. 1960-1992. One difference is that during that period, it was the Democrats who were outperforming their vote share.

2. The Effects of the 2010 Gerrymander Are Overstated.

Trende writes in another article, How Mich. Rebuts Redistricting/Polarization Claims that despite the overwhelming success of Republicans in that state, the district lines changed very little from 2010 to 2012. In fact, the Republican success came in 2010, before the lines changed. They just held on in 2012, despite no district crossing the 50% boundary in the Republicans' favor (i.e. no Democratic district became Republican). The only district that crossed the 50% barrier went from a small Republican majority to a small Democratic majority.

3. The Senate Is Also Polarizing

The Senate does not redistrict. Each Senator's district is statewide and states rarely change their boundaries. So gerrymandering can't explain polarization in the Senate.

4. Gerrymandering Is an Effect, Not a Cause.

In particular, Trende found

Under the present lines, using the McCain/Obama results, there were 123 seats that were R+10 or more and 110 seats that were D+10 or more. If we take the exact same lines, but use the Romney/Obama presidential results as our yardstick instead, we find 141 seats that are R+10 or more, and 121 seats that are D+10 or more. Looking at our broader set of Heavily Partisan Districts, we see eight additional heavily Republican districts and two additional heavily Democratic districts come into being just from the country becoming more polarized from 2008 to 2012. In a very odd way, we’re gerrymandering ourselves.

Even keeping the lines constant, as if redistricting did not happen, polarization is increasing.

Another reason that Trende does not mention.

5. Democrats are more likely to be in highly partisan districts.

The Cook Political Report creates a statistic it calls the Partisan Voting Index. The Cook PVI measures the presidential returns in the last two elections for each congressional district and turns them into a simple measure of partisanship. The 2018 list is PDF archive. For simplicity, I get this from Wikipedia, which is not paywalled or archived.

The point is that according to the Cook PVI, the twenty-one most partisan districts are not Republican but Democrat. And the majority of these happen not in states where Republicans controlled the redistricting process, like Ohio and North Carolina, but in states where the Democrats did.

• California: 6; a Democrat dominated commission drew the lines.
• New York: 6; Democrats controlled the House and governorship; control of the state Senate was split evenly.
• Illinois: 2; Democrats had complete control of the state legislature and governorship during redistricting; Republicans lost four seats after redistricting.

The remaining seven were each from different states, some of which were partially or completely Republican controlled. However, note that two thirds of the top twenty-one (fourteen) were Democrat dominated, even without including Massachusetts and Washington.

What do these twenty-one districts have in common? They were all in major cities: New York City; Los Angeles; Chicago; San Francisco; Oakland; Buffalo; Miami; Atlanta; Philadelphia; Boston; Detroit; Seattle.

There are multiple measures of gerrymandering, which have different bases and which may conflict:

1. Geographic compactness. A district should have as small a ratio between its border length and its area as possible.
2. Political compactness. A given political region should be all in one district. E.g. a city smaller than a district should be all in one district.
3. Proportionality. The districts should be divided in a ratio as close to that of the voters. So if 40% of the voters are Democrats, then 40% of the seats should be. This can be problematic in that votes change every election while the seats often change only once every ten years.

We see some confusion about proportionality in particular. Note that this does not mean that if a given election goes 52-48 for one party that the party with 52 should win a majority of the seats. It means that the seats should be divided in such a way that the voters are split in reasonable proportion. There are multiple reasons why vote share is not a good proxy in United States elections:

1. Not all seats are contested. A seat where one party wins 100% of the vote because the other party does not even run a candidate is not a 100% partisan seat. It may not even be an 80% partisan seat.
2. Some candidates are better than others. If one party runs a great candidate, that candidate may outperform the partisanship of the district.
3. Some candidates are worse than others. Conversely, some candidates run even though they have no chance. They raise no money. The incumbent may not bother to debate. Why should that candidate's inferior performance count against the party? Should a party have to run a superior candidate just to make the statistics work? How would a party reward such no-chance candidates for running?

To get more proportional results in these partisan districts, we would need to make major changes. We could throw out compactness and split these cities into multiple districts that each have urban, suburban, and rural portions. We could gerrymander the heck out of suburban districts to counteract the natural partisanship of the urban and rural districts. Or we could just change the elections to be some kind of proportional rather than having districts at all.

You may be asking about the twenty-second district. Is it also a city? No, it's a sprawling rural district, one of the larger districts in the country. It's overwhelmingly Republican as are all its neighboring districts (e.g. the twenty-fourth most partisan district is adjacent to it). Without drawing it as long, skinny (not compact) districts, there is no way to include it in significantly less partisan districts.

Exactly half the districts from twenty-two to thirty-one are Republican districts. But from thirty-two to forty, they're all Democratic again. Of the forty most partisan districts, five are Republican. Thirteen are in Republican controlled states (including four of five Republican districts). The exception was a Republican state with a Democratic governor. This is not what we'd expect to see if gerrymandering were the issue.

This is what we'd expect to see if Democrats are actually more polarized than Republicans. Urban areas are extremely Democratic. More extremely than rural areas are Republican.