Gerrymandering

Key Takeaways

a political party can earn votes through persuasion well-liked policy and high-quality candidates political parties can also win elections by redrawing districts in a process called gerrymandering in which voting districts are redrawn not with the intent of collecting geographic communities together but rather to strategically position boundaries of support this is data skeptic consensus in the 22nd installment in our series about how multi-agent systems achieve collective decision-making today on the show i speak with brian brubach for a deep dive on these topics and a discussion around his recent paper meddling metrics the effects of measuring and constraining partisan gerrymandering on voter incentives i'm brian brubach and i'm an assistant professor at wellesley college and tell me a little bit about the type of work you teach and research there i'm actually brand new as a faculty member there now which has been kind of crazy during the pandemic but i teach everything from introductory programming all the way up through algorithms and computer science theory which is more in my specialty as well as courses involving the social implications of computing which i think we'll talk about a little bit today and then my research is fairly broad ranging in terms of mostly algorithms in theory but with a lot of different applications from e-commerce type models to bioinformatics and i guess what we'll be talking about a lot today which is going to be more democracy redistricting gerrymandering fairness things like that now when you mention algorithms i'm afraid i may have biased the audience by giving unfair well i don't know if it's unfair but quite a bit of time on the show to machine learning algorithms specifically an algorithm of course is just a procedure but could you talk a little bit about some of the tools machine learning or otherwise what are the methodological toolbox pieces you use when you work on algorithms so initially i did a lot of work on intractable problems a lot of approximation outcomes to np-hard problems or problems that are hard in other ways i'm not sure how technical you want to get here things like formulating problems as a linear program and then using lp rounding techniques to get some good approximation so that's been a lot of the work that i've been doing it's involved you know some kind of randomized rounding techniques i can go into more detail on what that means or why that's necessary yeah i'd love to dig a little deeper something that's fascinating to me is that connection between social sciences and theoretical computer science these things that often seem very distant and may not even be in the same departments at a lot of universities yet there are connections there what does that look like in theory we could talk about it from a number of different directions in one case there's just the same theoretical foundations that you would look at for any kind of problem so there's theoretical foundational work on networking problems that's not necessarily immediately tied to real clear networking applications similarly we can look at theoretical bounds on how fair however you want to define that we can make an algorithm so there's that kind of theory work where i've done some work in fair clustering where you're maybe trying to balance demographics of a cluster or in the case of one of the papers we might talk about today if you think of clusters as districts you might have other fairness criteria that you're trying to preserve when you're making the clusters in one sense i think we'll probably talk more about this but i like really thinking about in terms of things like redistricting what can we prove is even possible because there's some things you might want to guarantee for a voter say in terms of a right that a voter has in an election and if you don't have any method for drawing a district that gives them that guarantee then you can't give them that guarantee so i think it's interesting in that regard to just look at what's theoretically possible along those lines so that's been some of my work and could we delve into the word fair a little bit i can't imagine you could find a reasonable person who says they're against fair algorithms their pro-unfair algorithms or something like that but i imagine there's not a consensus on what it means to be fair are there popular ideas or common approaches for measuring fairness there are and it's one of the things that gets a little bit thorny when you start talking in terms of theory versus applications in many cases in theoretical work the word fairness will get thrown around with a very specific definition you know some very specific mathematical definition for a problem so for instance if you're talking about a clustering algorithm that's clustering data points where the data points correspond to people you know you might have something called demographic fairness where you want the demographics of a cluster to match the demographics of the population but that's only one notion of fairness but it's a notion of fairness that has a very clean mathematical definition so sometimes that'll get thrown around in clustering it's just this is what it means to be a fair cluster but there's actually in any case there's going to be lots and lots of reasonable definitions of fairness based on what your actual application is so once you start getting down to real applications then you get more clear definitions of fairness so really there's tons of ways to define fairness for these problems it's really important to have that domain knowledge and know what fair means and so in the case of fair districting for example that's a really difficult thing to nail down and i don't think we have a really good idea of how to even think about that yet although many people i guess claim to some of the arguments we'll see in terms of even the more recent supreme court case on redistricting will look at the idea that maybe the whole process is intentionally designed to be unfair right that state legislatures should be allowed to draw what many people would consider to be unfair districts that that's you know what the constitution mandates just the idea of nailing down a definition of fairness for a specific application is in many ways you know a whole line of research is gerrymandering something that's relatively new or is it just new analytical techniques that are being embroiled in an old process i mean this is one of the things that's really fascinating and thrilling about working on some of these types of problems is that gerrymandering is very old i mean the term gerrymandering comes from 1812 and i guess technically there's somebody listening to this who's going to say it's gary mandering because the guy's name was elbridge gary but everybody pronounces it gerrymandering so it was inspired by a district back in 1812 so it's been around in that form in the united states since at least then but it's changed in more recent years in the sense that we now have computers that can draw these districts more effectively than say a person who was just in 1812 a person drawing a district was maybe going to be less skilled at gerrymandering than a computer with all the data that it has access to on past voting history can today i think one of the more extreme examples is north carolina where really you can redraw the maps now if you're using a computer to process the data and you can essentially get a 14 vote swing in the house of representatives if you think about flipping seven seats one way or the other that's a state with only 13 representatives i say 14 vote swing because when you flip one person it takes one vote away from one party and one to the other and that 14 vote swing that's larger than the difference between the majority and minority in the house currently so in recent years i guess it's become a more urgent issue as well as i guess political strategies involving trying to take over state houses to be able to redraw the maps more effectively but what's also really interesting is that a lot of the more rigorous ways of measuring gerrymandering have been very recent so a lot of these approaches that involve trying to kind of sample from the space of possible maps and show that some map is an outlier you know a map that is gerrymandered is an outlier among the space of all possible maps you could draw is a really new approach and it's kind of exciting to be you know thinking about a problem that's been around for over 200 years at least and just measuring this even though a lot of us could say oh we can look at a map and tell that it's gerrymandered it's usually not as simple as that a map can look gerrymandered and not really be gerrymandered or it can look fine and actually be gerrymandered it's only been in the past few years that we've had some really strong ways to quantify this that have held up in some state court cases and things like that if i were the villain of the story and i worked for one of the two parties political parties doesn't matter which one and my goal is to set fairness to the side and use any technique i can to redraw them in favor of my party winning what sort of data do i want to inform my decisions and how do i go about doing that i guess i would start by saying it's a little bit thorny to say you know you're the villain in the story even though i mean i can openly say that you know i don't want somebody doing what you're talking about doing and i think it's probably not good for somebody to do what you're talking about doing legally you're not necessarily the villain legally you're potentially just acting the way you're supposed to if it's known that you can gerrymander and that's part of the constitution then there's maybe some idea that that's part of the way the electoral process works but anyway setting that aside whether or not you're a villain which maybe i personally would say you're a villain in this case but ultimately we just want like a good representative democracy one way or another and so sometimes i think it's important especially when you're analyzing these processes try not to take a side when you're looking at just the effects and trying to really look at the long-term effects and i think in most cases we would say that the long-term effects are bad but as a computer scientist you know i like to just be able to show what's possible anyway so what you would want to do in this case is you would want to look at past voting data which is available in a lot of cases and any polling data that you might be doing internally with your political party and try to draw districts that are going to favor your party based on how you expect the voters to vote i think the simplest way to think of this if you google it online and look up pictures of gerrymandering where you can just assume let's say we know how everybody's going to vote and we want to draw this map and you're going to use two main strategies if we're talking about partisan gerrymandering there's other types of gerrymandering we're talking about partisan gerrymandering where you're just trying to make one political party win over the other your two main strategies are packing and fracturing packing refers to putting a bunch of people of the opposing party into one district so that their votes are essentially wasted if a district is 80 people from your opposing party then all of those votes over 50 are wasted and similarly if you want a fracture you might take that group of voters if there's a say a collection of like-minded voters from your opposing party you could split them up so that they're a minority in a bunch of other districts right you might have a group of voters from the opposing party that is more than 50 the size of a district and if they were all in the same district they would elect a representative of that party but if you split them in half or in multiple pieces and put them out into districts that otherwise favor your party then again you've kind of wasted their votes essentially you're trying to maximize the effect of your votes in your party and minimize the effect of the votes in the opposing party i'd love to talk about measuring gerrymandering in particular i think the two techniques you focused on were proportionality and outlier detection which i want to get into but maybe could we set the stage with a rough definition of compactness and maybe use that to contrast the other techniques against it what is compactness and why does it seem like it might be useful for gerrymandering compactness is one way that we can measure a district i think you're referring to geographical compactness we've all seen pictures of what is you know like a classic gerrymandered district and it's all long and scraggly a compact district would look more like a circle one of the common definitions would be look at the area related to the perimeter so that the actual formula is four pi times the area over the perimeter square that will give you a measure of compactness and in that case a circle is going to be the most compact shape that you could have whereas a long scraggly thing would have a very poor score by that method and that's something that has been used in some algorithmic techniques to you know say hey we're going to come up with a an algorithm to draw these districts and we're just going to value compactness and i think that that's valuable work that's one approach is to just say hey let's hand this problem off to an algorithm that's just optimizing for the most compact districts and not have legislators drawing the districts i would say two main problems with that well sometimes that's touted as a fair method it's not clear that drawing compact districts is unbiased it might be that a map with the most compact districts is biased towards say one party or the other especially when you have a situation like in the united states where one party is more common in urban areas and one party is more common in rural areas geographic measures could differ between those two parties and that's actually something that's been brought up in some of the supreme court cases and then there's the issue of you would have to get politicians to agree to that method which doesn't seem like it's going to happen anytime soon we would essentially be talking about changes to constitutions to say that instead of state legislatures drawing these districts that you would just have some algorithm draw them and that seems like something that a lot of people wouldn't want to give up the opportunity to do yeah there's something novel about it i don't want to discard it but i certainly wouldn't say it's sufficient to just solve the problem tell me a little bit about some other techniques why were proportionality and outlier detection better solutions for your work so proportionality is the idea that let's say we're talking about representatives from a state if a state is one-third from one party and two-thirds from the other party that maybe the representative breakdown should be the same you know one-third two-thirds and that's not our electoral system actually moving to that system would be a whole change in our electoral system even though you know a lot of times we look at a redistricting map that would give proportional representation is often seen as a fairer map and maybe it is you know if you want to take a bad districting map and go to the supreme court and say hey you know you should throw out this map you can't do it on proportional grounds because there's nothing in the law that says hey you should have a proportional map there's also issues like i'm calling in from massachusetts and there was some great work you know showing that you can't draw a proportional map of massachusetts because the members of the two major parties are basically living amongst each other it's not as segregated as some places and so it's not really possible to draw a map that would give i think it's 30 some percent of the massachusetts representatives to the republicans the rest of the democrats right there's every way that you draw a district map of massachusetts it just awards all of the seats to the democrats that's another issue with proportional methods and one reason we talked about them was also going back to the other thing you mentioned which has been successful which is these outlier approaches sometimes it's been said that restricting outlier maps is just a form of you know enforcing proportionality is a proxy for saying hey we want to favor proportional maps and the interesting thing you can show is that in fact banning outlier maps can sometimes ban a proportional map if a proportional map is an outlier it's certainly not as simple as it's sometimes been described and the idea with outlier maps is you know suppose you could look at the distribution of all possible maps let's say you looked at all possible legal maps of a state so by map i mean a way to draw the districts and you know in the simplest measure let's say you looked at the partisan breakdown of the seats assuming everybody voted the way you expected them to if the map that you've drawn produces an outcome that is very rare among that space of all possible maps you would say that it's an outlier and that would be one way to say that it's a gerrymander this didn't happen by accident this was you know a really intentional effort to you know maybe in some ways you could say disenfranchised people based on their political beliefs which is one way of phrasing it to get into the realm of violating somebody's rights we can't look at all possible maps obviously because that's a very large space it's not even possible to really sample from that space but what some people have shown is that you can approximately sample from that space in a way that's been seen as good enough from a legal perspective and again this isn't my work this is work from others that came for me but still in very recent years you're going to sample say 20 000 maps that are pretty close to a random sample of the space of possible maps factoring in things like the districts have to be connected and other restrictions like that and you're going to say okay there's 2000 possible maps and only 100 of them produce this very weird outcome that your map does so it's an outlier so you know you can't draw a map like that that was something that we took a closer look at in one of our papers where we said okay you know if that becomes a regulation in a sense either de facto or actually written in the law how does that affect voter incentives at least at a theoretical level the key thing to think about there is when you're doing this random sample of maps and you're looking at what the likely outcome is like you know maybe fifty percent of the time there's ten republicans and seven democrats or something like that but only one percent of the time are there twelve republicans and five democrats that's probably not an accurate example but you get the idea so you're doing that based on past voting data which is taken from prior elections and one thing we looked at on a theoretical level is how does this affect voter incentives because now your vote is counting in the election that your vote again but your vote is also being used later on to determine whether future maps are gerrymandered or not i think one of the most interesting things we came up with is maybe not surprisingly yes that does affect your incentive as a voter i'm not claiming that voters are going to go out and change how they vote because of this because again we're talking at a theoretical level right now but theoretically it does change your incentive as a voter and it can make it so that even in a two-party election right so let's say you know democrats and republicans in the u.s you can vote for the opposing party let's say there's two candidates running in your district you might vote for the candidate you like the least and that could be a good strategy for you which is usually not the case usually we only think of that happening when there's three or more candidates maybe you prefer some third party candidate but you vote for your lesser of two evils among the top two candidates right right because you believe they'll have a chance of winning yes exactly you vote for the candidate that's most likely to win the one you like the most of the ones that are likely to win that's the classic thing that we see a lot and so this is kind of fun and that it happens even when you're voting between two candidates and so people listening who might be familiar with the famous gibberd satterthwaite theorem we'll say okay well how does that happen and i don't know if you've covered that in the podcast i guess it's worth just mentioning it real quickly either way the basic idea is that if you have an ordinal voting system where you're electing a single candidate you're having an election where people have ordered preferences over the candidates and you're electing one candidate there's three possibilities one of which is going to be true one is that your system is a dictatorship which is basically saying one person just gets to pick the winner the other situation is there's only two candidates and the third situation is that the system isn't strategy proof meaning there's tactical voting right so again we've all seen if there's three candidates you might want to strategize and vote for the candidate that isn't your top choice and so the key thing that's different here is that that theorem works when you're electing a single candidate but here because the votes are being used in determining later regulations and your vote is potentially affecting a future election your vote isn't necessarily tied to a single election electing a single candidate so that's why the theorem doesn't hold in this case and that's why you have this really bizarre behavior where you could potentially want to vote for your least favorite candidate out of two and if you look at our paper there's actually a nice little picture you can look at this on like a three by three grid where you can see how essentially voters could flip their votes to make a later redistricting map look fair so you add a vote of the opposing party you're gonna flip your vote in an election that doesn't matter if you vote for the opposing candidate but your candidate still wins but you haven't lost anything but your vote for the opposing candidate could affect whether a future map looks fair or not and so again to actually implement this in practice you would need a large number of people coordinating and i'm not sure the practicality of that well i don't know either but i know the internet does make it possible for large groups of people to coordinate yeah i'm certainly not going to rule out anything but yeah so we looked at a way that you could potentially strategize around this regulation and just in a theoretical social choice theory setting how these types of regulation or anything that's going to involve past votes might affect voter incentives well let's get into some of the specifics of the game one of the reasons your paper caught my attention was it was the first sort of game theoretic treatment that i'd encountered that really delved into some of these voting choices if we think of it as a game what are the players doing and what are some of the strategies they might adopt i'll try to explain this in a way that somebody at home could even play this a picture truly is worth a thousand words so seeing the paper helps but yeah yeah this is actually what we did when we were first talking about this you know we sat at a white board and played this game against each other my co-authors and i imagine just a three by three grid of squares you can almost think of a tic-tac-toe board and you're going to want to divide that into three districts of three squares each and you want those districts to be connected meaning all of the squares in a district have to be connected by touching on the sides not on the corners and there's actually 10 ways to do this on a 3x3 grid so that makes it a little bit easier there's only 10 possible maps you can draw on this 3x3 grid and 3x3 is good because you have an odd number of districts you have an odd number of voters in each district so you don't have any ties so now imagine that your two parties let's say are x's and o's to keep with the tic-tac-toe thing let's say the first row the top row is all x's and then the bottom six squares are all o's if you look carefully at this you'll see that there's one map you can draw that will give the o's all three districts if you draw basically a map that is three columns where there's always two o's and one x and a district then the o's are always going to win all three districts and they're going to essentially send three representatives but the other nine maps out of ten the o's will only win two of the districts so you could draw this map either by yourself or with a friend and you can explore this game you could also put the x's in different spots but that's one way to think of it that's how we have it in the paper so what's gonna happen is let's say that on this map in particular if we're gonna say that a map is an outlier if only one of the ten maps produces the results right so only one of the ten maps produces three districts for the o's so we'll say that's an outlier so you can't draw that map you're regulated away from drawing that and the game's going to go in four rounds so the first round the majority party which we'll just say is the o's to start with will assume that they control the state house at the beginning is going to draw a map subject to that regulation so they can draw any of the nine maps that give the o's two districts and the x is one so that's round one the majority party draws a map round two all of the voters are simultaneously going to vote and they can vote for either party they have a preference but they could theoretically switch their vote and to make it simple let's just say that each party has total control over their voters which i think is clearly not true but we'll simplify it for this example and say that's the case and so everybody's going to vote simultaneously those votes are going to be tallied and whichever party won the majority in each district is going to get a seat from that district then in round three whichever party holds the majority most ways if you do this is going to be the o still whichever party holds the majority after round two meaning they won at least two of the seats is going to draw a new map again subject to the regulation but this time they get to look at the votes from round two so we started maybe seeding with some original voter preferences the x's along the top and the remaining o's but now they're going to look just at the votes from round two and draw a new map and they just have to make sure that that new map satisfies the outlier criteria it's not an outlier map then in round four you're again going to all vote simultaneously and since we'll say in this simple version of the game that this is the last round we can assume all voters just vote their true preferences and the goal if you're the o team in this is to win three seats in that fourth round to draw a map in round three that looks like it only gives you two seats but then when you actually vote your true preferences in the fourth round you win three seats and what we actually show in this simple example is that there's actually nothing the x's can do so the o's can vote strategically in a way that the exes can't counter that strategy in any way the o's can always win a total of five seats over the course of these four rounds that's the basic game and you can imagine that happening over more than four rounds but the basic idea is that you have a round of voting and then you have a round of redrawing the districts based on those votes and when you set it up that way we show that you can find strategies for getting extra seats for your party i think in this case you're only getting one extra seat and i think in practice that's probably in many cases the best you could do so for example if you had this regulation might still control a lot of gerrymandering well i shouldn't say that because again i'm not going to say that this is going to happen in practice but even if this did happen in practice i doubt it would have more than a one seat swing but a one seat swing can be fairly substantial when you consider that say right now once the last two seats are resolved i don't know when this recording is gonna post it's probably gonna be about nine seats apart and flipping one seat would take it down to seven seats apart so again these minor changes can have large effects so you'd mention the need or the strong assumption that the political party controls all of its supporters which i agree with that would be a very strong assumption but could this surprise attack come in another form could it be some fraud system or some hacking attempt that uses a feature like this to manipulate the system and you know while i guess making the least number of surreptitious vote changes for the maximum downstream effect something like that yes so this model that we look at in our paper is kind of the first work looking at this and we've kind of chosen the simplest possible model where each party has total control over their voters and i think it's really interesting future work to look at other models which could include fraud like you're talking about it could include a model where rather than just being able to decide how every voter votes you're spending money maybe you choose to just not spend your money on get out the vote in some areas and you choose to spend it in other areas or something like that there's lots of more complicated models that you could explore certainly fraud is one thing to talk about i guess i want to be very very careful right now how i speak about voter fraud but i guess the thing that we mention in our paper is that you could theoretically help your party by suppressing votes from your own party by switching votes from your party to the other party in a calculated way i'm not saying that there's a mechanism to do this or that this is likely to happen but were there to be a situation where an election could be compromised in that way that is something that could be done so again for that to happen you would have to have this regulation in place which it's a de facto regulation i guess in some states where it's been successful so in pennsylvania this approach was successful in overturning a district map so it would have to be a place where this measurement technique was being used as part of the criteria for determining a map to be gerrymandered and throwing it out and there would have to be i guess incentive to do it and i think theoretically you could change votes from your party to another party which would be kind of a weird thing to detect and that's why we mentioned it in the paper so if you were looking for voter fraud you might be looking for a party giving themselves votes not giving their votes to the other party and we're just showing that under this model of regulation you could theoretically pick up a seat with a better more favorable map by switching some of your votes to another party but i will say developing such a strategy is also very difficult maybe that's something we could touch on that we haven't talked about yet which is that in order to do any of this you also have to identify which voters should switch their votes and that's very much a non-trivial problem yeah and how you get them to collude and inform them of what they need to do and that they follow through would take maybe a grand conspiracy but the mechanism exists i guess maybe is the point yes and even if a party has total control over its voters say they would still have to solve the difficult computational problem of identifying which voters vote should be switched so that they don't accidentally cost themselves some seats in the current election and so that they can also draw their favorable map from the future and they would have to factor in you know that the other party could be doing the same thing and i'll note that a lot of our work is on simple models like these grid graphs like a 3x3 grid or a 5x5 grid and when we extended some of this our heuristic for finding these strategies to real north carolina data we used a restricted model where we were assuming that the other party isn't also strategizing because it was just too difficult at least from where we were when we were doing this paper to do the full computation of considering everything that the other party could do and what is the computational challenge is it a bit like chess where you're exploring a big decision tree of moves or something what is the computational challenge yeah that's probably one of the simplest ways to think about it is just that there's a huge decision space and you can prune away some of that but if you imagine the thing that you can do here is for each individual voter you can tell them which way to vote so that's two options and so when you have millions of voters that's two to the one million possibilities and you can prune away that space but you're still going to have a very large space probably which you would have to do in practice if you were really thinking about this is something called a mixed strategy where you have some probability of flipping some of these voters and some probability of flipping some of those voters the only way that we could actually do this on the 5x5 grid where we were able to have both sides strategizing is that we were able to find in that restricted setting these pure strategies where the majority party can do something and there's nothing that the minority party can do and i guess the even simpler example is the one in the paper on the 3x3 grid we show that there's nothing that the minority party can do and so that makes it a lot easier to find a strategy for the majority party because we don't have to consider all the things that they might do in a real district map that's not going to be the case in a real district map there is going to be probably something that the minority party could do or that the opposing party doesn't necessarily have to be a minority but the posing party could do and factoring in all those different strategies it just becomes very computationally intensive and it gets into the realm of solving large games well we've talked about a lot of ways in which you know under these terms things work out this way or under a lot of situations i guess we're sensitive to the conditions we choose the rules of the game and that sort of thing so these insights are novel from a theoretical perspective do they also give us any insight into the country we live in the united states if i know some congressional committee reached out to you and said you've been thinking about this more than we have can you tell us what our priorities should be for addressing gerrymandering if anything one thing i would say since we've been talking about this paper of how you could subvert the outlier regulation is i still think the outlier regulation is overall a good idea and it's been a good approach that has yielded some good results so i'm not in any way in this paper saying that that's a bad way to measure gerrymandering we're just looking at how it affects incentives and possibly how people could try to subvert it at least partially i do think that that based on everything that we know at this point that the method of looking at outliers and throwing away outlier maps is probably a good thing to do i also think we really need to rethink what it means to have a right to vote and what right a voter should be guaranteed so one thing that can come up here when you look at the idea of okay we sampled 20 000 maps and we can see what the outcome of election would be in all of these maps you might say well if a voter is in a district where their candidate wins in the majority of maps do they have some right to be in such a district one thing we show is that you can't actually give that guarantee to every single voter might be that there's no single map that gives every voter who has that quality that guarantee that says every voter who gets their choice in a majority of maps also gets their choice in this map there's scenarios where that can't exist but maybe there's some definition of fairness that you can use some right to be able to vote with their neighbors i come from a randomized algorithms background so i'm predisposed to the idea of a randomized districting i think that that probably would rub a lot of people the wrong way but for me i think it's one of the cases where the most fair thing you could do was something that involved randomness i'm not saying you just randomly draw the districts but some distribution on possible districts would probably be the most fair give people the most chance to say vote in the same district as their neighbors or as people who are like-minded to them and essentially give them the most possible right to vote because in a deterministic system right there's going to be people who their vote essentially feels like it never matters right if they're always voting in an election where their candidate doesn't win and if you can randomly draw districts such that maybe 20 of the time their candidate wins i think that that's an interesting thing to think about i think it's maybe outside the scope of my expertise to really say whether that would ultimately produce a good representative democracy but it's something to think about and i certainly would in some of my work i'm hoping to just show what is possible in these regards like suppose somebody says it would be good to draw this kind of district or to draw districts in this way or following this criteria say of randomly drawing them to show that such a thing is possible computationally is something that i'm very interested in because again i think that some of these things that we would like to offer in terms of voter rights you know we can only guarantee them if we can actually show that we can draw the districts this way and so looking at that as well as just saying hey whenever we do stuff whenever we make decisions here let's look at the downstream effects let's look at how this affects voter incentives let's look at how people are going to try to strategize let's try to stay a little bit ahead of all of the people whose job it is to subvert whatever rules we're making and analyze that as best we can well brian thank you so much for coming on the show to share your research and expertise thanks for having me that concludes this installment of data skeptic consensus our guest this week was brian brubach claudia arm brewster is our associate producer vanessa burciaga does guest coordination and i've been your host kyle police [Music] you