Steven Cherry Hi, this is Steven Cherry for Radio Spectrum. Today's episode looks at a collision of mathematics and politics.
In the 2020 elections for the North Carolina State House, Democrats received 49 percent of the votes but won only 42.5 percent of the seats. In three-quarters of the state-level elections, the winning margin was more than 20 percentage points—in other words, landslides—even though statewide, the margins between the two main political parties is razor thin—at the presidential level, Trump beat Biden by less than 2 percent, and a Democrat won the 2020 governor's race.
That's gerrymandering, the process by which a state is divided up in such a way as to maximize the number of electoral seats one particular party is likely to win.
And that was after a 2019 state court ruling that required the legislature to draw fairer election districts. And indeed, earlier election margins were even more extreme. After the 2010 U.S. Census, Republicans drew 10 overwhelmingly Republican districts and three serpentine Democratic districts. The word “serpentine" isn't used by happenstance, by the way. The word “gerrymander" is a portmanteau that's almost 200 years old, combining the name of then-Massachusetts Governor Elbridge Thomas Gerry with the word salamander, which was said to be the shape of the one of the districts in his state redistricting plan of 1812.
Since 1790, every 10 years, the U.S. conducts a census of all its residents, as mandated by the U.S. Constitution. The Census is relied on by government and commerce for everything from the Federal Reserve's economic forecasts to where Amazon should build its next warehouse. But the original purpose was in apportioning the U.S. House of Representatives districts fairly among the states. It quickly came to be used to draw House districts within a state, and the districts for state legislators as well. Thus in every year that ends with an 11 or 12, in 50 state legislatures, whichever party is in power gets a chance to try to cement its majority by jerry-rigging, pun intended, its district borders.
There are two ways to gerrymander. In one, you concentrate your opposition's likely voters into a single district, giving that one away but winning all or most of the surrounding areas. In the other, you divide a concentration of likely voters into two or more districts in such a way that they'll fall short of a majority.
Gerrymandering is obviously unfair, but creating fair districts is harder than it looks. There's an absolute requirement that districts be contiguous. And ideally, they're fairly compact, but also respect logical boundaries such as roads and town or county borders, and natural features such as rivers and mountains. But to what extent should a redistricting scheme match the current ratios of registered voters of one political party or another? And to what extent should a redistricting scheme create districts that match the racial and ethnic ratios of the state as a whole—especially when people of one race or ethnicity or political affiliation are often concentrated in individual cities or counties? And should it maintain any of those ratios at the state level, or, to the extent possible, within each district?
So political operatives and consultants draw up various maps, maximizing this or that, but mostly their party's interests. If this seems instead like a job for computer-aided statistical analysis, it is. Several years ago, researchers in North Carolina got the idea of generating thousands—even tens of thousands—of maps, and creating algorithms that maximize the desired variables to the extent possible. My guest today is a leader of such efforts, in mathematical theory—and practice, in the form of serving as an expert providing analysis and testimony in a number of gerrymandering court cases.
Jonathan Mattingly is a Professor of Statistical Science, and a Professor of Mathematics and Chair of that department at Duke University He leads a group at Duke that conducts non-partisan research to understand and quantify gerrymandering.
Jonathan, welcome to the podcast.
Jonathan Mattingly Thank you. Thank you so much for having me here. I'm excited to have this conversation.
Steven Cherry Jonathan, there's a big North Carolina case. Rucho versus Common Cause. A 2016 congressional map drawn up by the state legislature was struck down in 2018 by a district court as gerrymandering. The head of the North Carolina Senate's redistricting committee, Robert Rucho, appealed to the U.S. Supreme Court, which heard the case in March of 2019, and that June ruled five to four, basically that it couldn't rule—that partisan gerrymandering was a political matter and not one for the courts to sort out. We'll come back to that ruling. But first, let's talk about gerrymandering. In a blog post, you wrote: This effect was very pronounced in the North Carolina congressional maps used in 2012 and 2016. The enacted maps produced the same result over a large range of elections. What does that mean?
Jonathan Mattingly It refers to the fact that these maps were drawn to essentially—whether it was done intuitively or by design—but the effect was that they baked in certain results. That is to say that under a wide number of historical elections, you can look back and take elections for various candidates in various races and use them as a plausible election for the congressional delegation and over a large range of elections under many different circumstances. They produced almost the exact same results, always a 10–3 split, maybe sometimes a 9–4 split. But it was there was nothing commensurate with the swings in public opinion that we saw. So it was nonresponsive—the results were essentially decided before the election.
Steven Cherry In your testimony in Rucho, you describe the analysis your group at Duke came up with. You found that in some redistricting plans, but not others, a change in the vote between the two parties created a proportional change in the vote. That is, if two percent more of the voters voted for, say, Democratic candidates, the result in each district would change by two percent. Is that right?
Jonathan Mattingly I don't think we really emphasize proportionality. I think proportionality, unfortunately, is where many of these conversations start. And it's not I don't think it's right to think that way. I mean, you know, there's a lot of it, especially in the popular outrage. Like you said in your introduction, you know, this percentage of the state voted for the Democrats, but yet they only got this many seats. Right. There's nowhere in our system that's designed to be proportional representation. We're not a parliamentary procedure. We're not a statewide vote count. And yet people are surprised when the system doesn't produce proportional representation.
I think the question I'd prefer to ask is we have some set of rules and those rules imply a certain way in which the personal votes at the precinct level are translated into statewide election outcomes. The question is, did someone put their finger on the scale when they implemented those rules? For instance, our Senate is not proportional representation. Montana has many more votes than it should have based on its population. But we have a societal idea that we like the idea of protecting Montana from the whims of L.A. and New York City. Gerrymandering conversations start at the point of how the rules have been implemented. If we don't like the rules that we've set in place, that's a different conversation and maybe one we should have. But I'd like to make sure that people separate the two conversations.
Steven Cherry A Common Cause statement noted that [North Carolina] legislators freely and publicly admitted that their goal was to carve out and hold a 10 to 3 advantage in U,S. House seats for their own party, despite the fact that the votes cast in those races would split nearly down the middle.
Jonathan Mattingly Yes, that's correct. I mean, they said on the record in the state General Assembly that the only reason they made it 10–3 split was that they tried to make an 11–2 split and they couldn't. When we look at the properties of those maps that they produced for the 2012 elections and 2016, they're different maps, they look very different. But politically, they're equivalent. They basically lock in over a wide range of voter opinions, voter expressions of their will, their political will. They lock in the same political outcome, a 10–3 split or close to it, a 9–4 split with one of them right on the edge.
Steven Cherry And if I understand it correctly, your methodology is to generate an ensemble of redistricting plans—in one analysis, over 24 000 of them—and then compare them to actual plans. The legislature's actual plan, for example. Then you have some graphs where the legislature's plan shows an enormous jump while other possible redistrictings do not. What's the significance of that jump?
Jonathan Mattingly That jump is ... in Rucho we called it the signature of gerrymandering. It's essentially the effect is exactly this non responsiveness, the fact that if you shift the political outcome up or down a little bit, you don't see a change in the political outcome. If you look at those plots, only one of the dots crosses over the 50 percent line, that is to say, switches from being majority Democrats to majority Republican, does the seat change hands. On our website, also on that same blog, we also have some movies which I think are maybe the most compelling way to explain what's going on. Unfortunately, movies are hard to show on the radio. Still, maybe that's in our future, but not yet.
Steven Cherry We can link to them, though, in the transcript. There was an additional line in your graph marked "judges." What did that represent?
Jonathan Mattingly Right. So an interesting thing. So Tom Ross at the Volcker Alliance for Better Government—I think I got the exact name right. He was president of the University of North Carolina at Davidson College and a former judge—ran a simulation. And I don't mean computer simulation.
There was a law that attempted to set up a bipartisan redistricting commission. And he got a number of retired elder statesmen of North Carolina, retired judges, many of them retired Supreme Court judges, five Democrats, five Republicans. And they made up a map. They argued about it. They tried to be fair. And that map, the dots from the judges map was their map.
Steven Cherry So where do we stand in North Carolina? There were, after the Supreme Court case, some other cases in state courts.
Jonathan Mattingly So of course, the North Carolina case went to the Supreme Court. And there, as you said, the Supreme Court in some way said it wasn't their business, but they also said that it might be the state's business. So in some ways, they kicked it back to the states. The North Carolina Constitution has an interesting phrase in it. It says that elections should be fair (the federal constitution does not). And so there were cases brought that appeal to the North Carolina Constitution in particular. And there was a first case, Common Cause versus Lewis, which I also was an expert witness in, and that took on the maps that were used for the state legislature.
That is just to elect our state house in our state Senate. We have a bicameral legislature at the state level also. And then after that case was won by the plaintiffs, so they were forced to redraw the maps, there was a second case, Harper versus Lewis, which took the evidence from—it was just months before the election had to get going—took the evidence from Common Cause versus Rucho, which you spoke about, which had gone to the Supreme Court, and resubmitted it, essentially. The transcripts of the case. And then the state court ruled, based on that evidence, that those maps were unconstitutional relative to the state constitution.
And so all three maps—the U.S. Congressional map for North Carolina and the two redistricting maps for the state legislature—were all thrown out and new maps were redrawn by the legislature for the 2020 elections. And you brought up that line that says the judges on it. There's a similar plot which may not be on our website, which shows that the new remedial maps are not as different from those judges maps as the previous maps.
The new remedial maps are not perfect. They again lock in an outcome which looks more like 5–7; 7 for the Republicans, 5 for the Democrats. But that's different than 10–3. So there was a shift and it doesn't lock it in quite as strongly—it seems from our analysis—as the previous maps.
Steven Cherry Let's talk about how hard it is to draw a fair redistricting. You wrote it in another blog post, "North Carolina's constitution requires that the state legislative districts should not split counties. However, counties must be split to comply with the one person, one vote mandate of the U.S. Supreme Court."
Jonathan Mattingly Yes, that's correct. I mean, so there's this tension. I mean, when you do this ... when people like to talk about fairness—and I want to somehow say part of the question here is what is fairness? We tend to take a very operational perspective on that and maybe we can come back to that. But to answer your question directly, there's a lot of peculiarities in each state and each state's laws, and you really have to take them into account. So if you think you have a broad sweeping idea of what should be fair and based on some magic, one line calculation you do that doesn't take into account all those peculiarities of local laws and local traditions, I think you'll have a hard time convincing a legislature or a jury, or a set of judges, a panel of judges of your correctness.
So that's one example. There's a tension here and there's a state court ruling which specifies for those ... We're on an electrical engineering podcast, so I'll just say it actually more or less lays out a greedy algorithm, a greedy optimization algorithm about what should be balanced in drawing districts in North Carolina.
Steven Cherry Maybe you should explain that for the non-EEs in our audience.
Jonathan Mattingly Ha! Yeah, fine. So by greedy, it means that there's a list of priorities and it says, first, you should favor this priority. First, you clump the counties together so that when you subdivide those groups of counties, you will stay close to the one person, one vote. And then once you've done that, you then subdivide each of the groups of counties as little mini-states, into their own districts, possibly splitting counties, thereby breaking the state constitution.
So in then it keeps going down different criteria. And then after that, you have to try to comply with the Voting Rights Act, which you talked about—racial gerrymandering. So there's a list of priorities. And you're supposed to do them in order; satisfy this first, then satisfy this, then satisfy this—as best you can.
Steven Cherry Yes. U.S. Supreme Courts in the past have ruled on gerrymandering cases, but those were cases of racial gerrymandering. From a mathematical standpoint, is there any difference between racial gerrymandering—that is maximizing the voting power of one racial group—and partisan gerrymandering between the political parties?
Jonathan Mattingly Right. And there are some differences. I mean, from a mathematical point of view. Racial gerrymandering first ask the question, is a group sufficiently large and does it have a sufficient group of like-minded people, maybe of the same race, maybe not to vote as a bloc that give a racial minority, which meets certain thresholds which are laid out the Gingles test [Thornburg v Gingles] from a legal point of view.
Did they have enough power to elect a representative? Then after that, the real question then is, did someone in satisfying that requirement overdo it? That's what the 2012 maps were declared. That they said, well, you tried to satisfy the Voting Rights Act, but in doing so, you put so many African-Americans in one district, you actually diminish their power. So that was why the 2012 maps were thrown out. So there is some relationship. It's asking kind of did you follow the rules beyond what you needed to do in such a way that you are benefiting your party? That's the real question. And the rules are not: proportional representation. The rules are not: make every district as competitive as you can. The rules are about compact districts. They're about not splitting counties. Our general methodology is to lay out the basic rules and then ask what kind of outcomes would you typically see? And so we replaced some idea of fairness with some operational idea of fairness, which is are your maps very standard? Do they look typical relative to maps someone would have drawn if they didn't know anything about political parties? And if your maps don't, then it's very suspect that you put your finger on the scale.
Steven Cherry Jonathan, we should note that gerrymandering is hardly confined to one state or one party. New Mexico and Maryland are just two of the places where we've seen gerrymandering by Democrat legislatures. In fact, a Maryland case was combined with the North Carolina Rucho case in that Supreme Court ruling. Is what happened in North Carolina for the past decade singular or closer to the norm?
Jonathan Mattingly It's not singular. There are many examples. In fact, we've done an analysis in Maryland and Maryland has a similar type of non-responsiveness that I just described in North Carolina. They've locked in essentially a 7–1 Democratic majority, even though sometimes the Democrats might have lost a seat. And in fact, sometimes the Republicans would have lost the seat, but they both gave up something for stability. You might almost say that it's a gerrymandering to benefit the incumbents as much as any party.
But there are many states where there's gerrymandering, there are so many states where due to voter initiatives, there have been redistricting commissions that have tried to alleviate ideas of gerrymandering. But there are many, many states that do it. I will say that from my analysis, North Carolina has been right at the edge of the most extreme cases, however, and those more extreme, the extremism in North Carolina has real political consequences because North Carolina is such a purple state. As you said, we have a Democratic governor. That's not unusual. We've in the past regularly had one Republican and one Democratic senator. So it's hardly a state that is locked into one party or the other.
Steven Cherry There's movement in some states to create nonpartisan commissions that would do the redistricting instead of the legislature directly. To what extent are those commissions likely to use statistics in the way that you have?
Jonathan Mattingly I think there's some interest now. I know that some of those groups have thought about it. I think that it's ... the design ... It's a very interesting mathematical question in its own right, the design of the procedures for running such a commission so that it maintains fairness and they're not all the same. They're very different. It's not one-size-fits-all. Right?
In Arizona, there's an idea of making districts competitive, which is complicated because when political opinions change what it means to be competitive changes. On the other hand, California tries very hard to identify and preserve communities of interest, which itself ... It's defining is a complicated conversation. But I think there is some good that could come from our methods saying here's what you would typically see and maybe when you're done, you should have chosen a redistricting that is within the signposts of normality of our ensemble of districts. It gives some idea of how votes typically translate into seats in your state, given the strange geography of your state, the strange distribution of people. When we talk about statewide vote percentages, it callously ignores the fact that people are clustered in different parts of the states and political opinions are far from uniform across the states, especially in a state like North Carolina.
Steven Cherry It seems that at the federal presidential voting level, the Electoral College locks in something that looks like gerrymandering into constitutional law. That is by apportioning two votes to each state, even though registered Democrats are clustered much more tightly than Republicans—in big cities and a handful of states and so on—the Republicans have had and will for some time continue to have an advantage that looks a lot like the packing version of gerrymandering.
Jonathan Mattingly I hear your question and I agree with all of your points, except that I might say that it's exactly a great example to emphasize what we ... what I mean by gerrymandering. I would say it's exactly not gerrymandering. We have a political system. We have some rules, and those rules are implemented in a nonpartisan way. And the outcome is the Senate, right? It's very straightforward. Every state gets two. That's the rule. And there's no room to mess with that. No room to modify it.
If we don't like those rules, if we don't like the political values that it represents, if we think that it's not right that some states have such a diminished political power in the Senate, that we should change the rules. And so I would like to make the clear separation that gerrymandering is about: Here's the set of rules that we've set up. Have they been implemented in a nonpartisan way or has someone put their finger on the scales beyond the rules to benefit their party?
For instance, in Wisconsin—we did an analysis, and in North Carolina, we did the same analysis—it's very clear that there is a clear bias, a clear advantage if your baseline is proportional representation, which it shouldn't be. But if that's where you came to this conversation, there is a clear structural advantage to the Republican Party because their votes are spread out geographically where the Democrats are more concentrated in the urban areas, although not exclusively in North Carolina, for instance. But that gives them a structural advantage. They almost always will get more seats than they get statewide votes—the Republicans do. However, that accounts for a certain level of tilt to the Republican Party. And what we saw in North Carolina and what we saw in Wisconsin was an extreme tilt beyond that.
So gerrymandering is saying here's what you should have seen given the rules we have. But yet you produced a much stronger bias to your party than the rules would have typically shown. Now, we may, as a community, decide as a culture decide that we don't like even the original bias and we want to have some multimember districts. Maybe we want to have some other form, rank-order votes, or some system to try to alleviate that bias. But the flip side of that is equally complicated, right? Multimember districts have a long history of being used to suppress racial minorities. So these conversations are complicated. There's always both sides to the coin.
I like to point out to people that I agree that it may bother you that Rhode Island and Montana have so much power in the Senate, on the other hand, you could imagine a system where we have a representative ... a perfect proportional representation system across the entire country and New York and L.A. decide to dump all their waste in Montana. And that would be a different problem because they would have the political power to do that. And you see that in other countries where the systems are very, very centralized, where the capital cities in the big urban centers rule with an iron hand over the entire country. So it's a complicated conversation with two sides.
Steven Cherry Yeah, it's as complicated as the meaning of the word fairness itself. Gerrymandering is a hard problem mathematically and politically. The mathematics, it seems, have been largely overcome. With the 2020 Census completed and nationwide redistricting almost upon us, are you optimistic or pessimistic on the political front?
Jonathan Mattingly We don't have a uniform playing field across the country, right? There's still a lot of room in our states for gerrymandering. Many states don't have the ability to put ballot measures motivated by the people. There's overwhelming support in the public for bipartisan redistricting initiatives that take the power out of the legislature or at least clip the power of the legislature. I see that's going to have a hard time getting through in states where there aren't ballot measures. That's one point.
Secondly, yes, we've gotten better at addressing these questions in court cases now. But remember, we won multiple times in North Carolina and yet the entire decade in North Carolina, every single election until the one in 2020, was held under maps that were eventually ruled unconstitutional. That didn't negate any of the laws that were passed during those periods. There was still an entire decade, essentially, where the legislature that was elected—both to Congress and to the state legislature—was not one that was, in retrospect, deemed constitutionally representative. We have to find a way better than every time going to court.
And I will say that there are still many mathematical challenges to overcome. I think we have a good new perspective on it. This ensemble method gives us ways to think about it, but we also need to find good ways to explain it politically to legislatures and to to the public in general. And there are still challenging methods. And some of these gerrymanders were so egregious that it was not so hard to see the one tree left in the prairie when everything else was cut down. But other situations may not be as obvious, and we may ask for more nuanced analysis and those require more mathematical investigations. This is kind of a classic example of a good problem, leads to beautiful mathematics, and I still think we'll see lots of beautiful mathematics in the next decade to make our understanding even better.
Steven Cherry I think it's very generous of you to put it that way. I mean, after all, you've worked with high school students to produce some of these analyses. If high school students can understand the math, it seems like politicians and judges must be actively looking the other way not to see the gerrymandering that's going on.
Jonathan Mattingly Yeah, I hear you. Although I will say that in North Carolina, every three-judge panel, constitutional panel that looked at these cases has always ruled in favor of the plaintiffs—that is against the legislature. Those panels were not predominantly one party or the other. They were always ... They were panels that had both Democrats and Republicans. And often the swing vote was a Republican who sided with the majority. So I'm not so sure ... There's hope in that respect. And I'm hopeful that there's enough political energy now and public opinion is strong at this point that they will move against gerrymandering. Right. Arnold Schwarzenegger and Eric Holder are together on this. So that already says something.
Steven Cherry So it's like the brothers in [the movie] Twins. Well, Jonathan, the political differences in the U.S. become chasms when one party can exercise minority rule. Gerrymandering is just one part of that. But I don't think it's an exaggeration to say that efforts like yours and your groups are an important contribution to the maintenance of democracy itself. Thank you for that work and for joining us today.
Thank you very, very much for having me.
We've been speaking with Jonathan Mattingly, Professor of Mathematics at Duke University and leader of a non-partisan group that researches quantifies gerrymandering.
Radio Spectrum is brought to you by IEEE Spectrum, the member magazine of the Institute of Electrical and Electronic Engineers, a professional organization dedicated to advancing technology for the benefit of humanity.
This interview was recorded February 16, 2021 using Zoom and Adobe Audition. Our theme music is by Chad Crouch.
You can subscribe to Radio Spectrum on Spotify, Stitcher, Apple, Google, and wherever else you get your podcasts, or listen on the Spectrum website, which also contains transcripts of all our episodes. We welcome your feedback on the web or in social media.
For Radio Spectrum, I'm Steven Cherry.
Note: Transcripts are created for the convenience of our readers and listeners. The authoritative record of IEEE Spectrum's audio programming is the audio version.
We welcome your comments on Twitter (@RadioSpectrum1 and @IEEESpectrum) and Facebook.