What is the reason that redistricting is not automatic? Is it just a legacy of a time when it was impossible?

The definition of gerrymandering is:

In the process of setting electoral districts, gerrymandering is a practice that attempts to establish a political advantage for a particular party or group by manipulating geographic boundaries to create partisan advantaged districts.

It occurred to me (mostly as I am a CS major) that this is a process that can be done automatically, via clustering algorithms with the goal being to minimize the distance between points in the cluster, and have the clusters be the same size. This is not a "hard" problem to solve (technically speaking) and would remove all of the bias that gerrymandering puts into the political process.

So, is there a reason this is not done?

  • 1
    There are several problems with this question: (1) Assumption that "academics" are non-partisan; (2) Assumption that gerrimandering somehow uniquely benefits whoever controls the statehouse (instead of pretty much all incumbents with rare exceptions); (3) That a clustering algorithm can find an optimal solution (I'd not be surprised if such a solution is impossible algorithmically)
    – user4012
    Commented Jan 23, 2013 at 20:20
  • As an aside, whether gerrimandering is that big of an influence is a big question mark, at least judging by P.SE lack of answer to my question.
    – user4012
    Commented Jan 23, 2013 at 20:22
  • @DVK is that better? Additionally, while a precise (provably optimal clustering) may be impossible/impractical, one that is guaranteed to be within very tight bounds would seem to be possible (see here)
    – soandos
    Commented Jan 23, 2013 at 20:39
  • As best I can tell, this is a common risk of having a single winner for a district and wanting geographically-bound districts in the first place. Remove either of these requirements (For example, multiple proportional winners or random allotment to a non-geographic voting block); and the problem of gerrymandering goes away. Within the constraints of single winner geographic districts, better algorithms would still require existing power bases to be completely overhauled - and the day vested interests are willing to do that, we can get rid of other absurd constructs like the electoral college. Commented Jan 25, 2016 at 11:25

3 Answers 3


Historically, there are two competing principles at work in figuring out districts:

  1. You want about the same population in each district

    This principle was enshrined in US Court law in Baker v. Carr an Wesberry v. Sanders. Malapportionment is inherently unfair.

  2. You want communities of interest not to be divided.

    That said, there is something to be said for ensuring that there is a diversity of opinion on matters, too. It is easy enough to be "equal" with numbers in such a way that certain voices completely disappear from the discussion. This is equally bad, and also to be avoided.

    In the USA, for example, Vieth v. Jubelirer basically said that a state can have unequal districts, if the state has a good reason - like if it ensures diversity or allows more voices to be heard.

The first principle of gerrymandering, for example, can be algorithmically derived. It is possible, if the population were say, 1000 people, and there were 10 districts, to put exactly 100 people in each district. Grant you, with freedom of movement, that would hold for about 26 minutes, but hey - there is nothing says you couldn't have districts that exactly apportioned, say, the population on the date of Decennial census.

An At-large seat actually does this in some ways - by lumping everybody into the same district, you most accurately represent the will of all the people in the district.

But, what you miss is a community of interest. Imagine, for example, a three-district region consisting of 100 people, 45 of whom are ardent Yellows, 45 of whom are radical Browns, and 10 of whom are Chartruese. (Only the females can actually recognize those people, but I understand it is a color!). Chartrueses realize they will never be in the majority - and so they tend to vote for the proposals that benefit everybody.

Assuming they are geographically evenly divided, it is trivial to create 3 scenarios:

  1. Gerrymander it such that B is the majority in 2 and Y in 1

    District 1 - 20 Y, 15 B, 3 C, District 2 - 20 Y, 15 B, 3 C, District 3 - 5 Y, 15 B, 4 C

    Here, we see that Yellow will almost always will Districts 1 & 2, and B gets 3. In no case do the C voices get heard, and most laws will favor the Ys, since they have more districts than Bs. We pretty much all agree this is unfair - to "crack" the Bs is bad. Alternatively, you could stick all 45 Bs into one district and "pack" them there- same basic effect.

    This is truly gerrymandering for political advantage, and is the unfair case.

  2. Put the population is 3 equal districts

    Each district could have 15 Y, 15 B, and 3 (or 4) C.

    On the surface, this looks fair, right? Except, here's the problem. In every district, the Cs hold all the cards. Any 2 Cs can swing the district for that entire election. As such, 6% of the population determines the outcome. If 6 Cs collude, they can silence every voice but there own. There are dictatorships that aren't this bad!

  3. Make 1 majority district for B & Y, and 1 competitive district

    District 1 - 20 Y, 10 B, 3 C, District 2 - 10 Y, 20 B, 3 C, District 3 - 15 Y, 15 B, 4 C

    Here, you have communities of interests. District 1 will almost always be yellow, District 2 always Brown. This way, the Y & B's voices will remain, but there is still the swing District C. This is arguably the most fair, because all voices are heard, and even the "disenfranchised" packed voters (the Bs in 1, the Ys in 2) have their voices being heard in other districts.

Now, obviously, the world is more complex, but communities still fall into these patterns. Rural voters, for example, share similar values in many ways. And, their voices should, in a fair society, be heard in, possibly even in greater proportion then their numbers suggest. After all, politicians have to expend more energy to visit a rural area than an urban one. If voices were exactly equal, rural voters would be heard less.

Communities of Interest can be geographical (think Virginia's Eastern Shore, Tasmania, Bavaria vs. Lower Saxony in Germany), ideological (think any party), or based on other factors.

Balancing these factors is why humans get involved in the first place.

  • And I suppose that the fact that this leads to incumbents being very difficult to remove is not considered a bad thing?
    – soandos
    Commented Jan 23, 2013 at 22:54
  • 5
    @soandos: not by the incumbents who are doing the redistricting. Commented Jan 23, 2013 at 23:49
  • 1
    +1 Millikan :). In all seriousness, if you can devise an algorithm that can balance voices, all the power to you. So far, getting people involved seems to be best. Personally, I prefer commissions to legislatures, but you asked, why not just have equal districts. Hopefully this gives you that why. Commented Jan 24, 2013 at 1:00
  • Here in Ireland, we have multi-seat constituencies, usually based on traditional county boundaries dating back to well before the foundation of the state.
    – TRiG
    Commented Feb 1, 2013 at 20:55
  • If you group all the like-minded people into a single district, doesn't that encourage extremism? Also, how do you determine which like-mindedness is important? Should we group by race? By support of 2nd amendment? By feelings on abortion? By religion? Choosing any of those to group together means less grouping of the others?
    – Readin
    Commented Feb 14, 2017 at 4:24

You already know the answer: It is, indeed, "just a legacy of a time when it was impossible". Part of that legacy though is people not understanding what computers can do. I wrote automated redistricting software ( http://autoredistrict.org ) and I have been pushing for its use. One of my main difficulties has been people who didn't think it can do all the things a human can do. And the things they mention are usually actually really easy for a computer to do. Like measuring compactness, for example.

Another thing I think is humans don't want to lose power. And a lot of people seem to be under the mistaken impression that "communities of interest" serve those communities, and that it's ethical to have a human pick and choose them. Both are of course false. Communities of interest is another name for vote-packing and, like any vote-packing, it disfavors that community's preferred policies.

So, three things, all human errors: the mistaken understanding of what computers can and can't do, the mistaken belief that it's ethical for a human to pick voters and candidates, and the mistaken belief that things like communities of interest are good things. All a consequence of insufficient education and/or thinking through the problem.

  • This looks really interesting and I'm going to take a good look at it later today probably. Do you have any metrics on how well balanced the districts are by population and how it scales (i.e. could it do something like NY on a census block level with < 1% population difference between all districts and the ideal population level per district)?
    – soandos
    Commented Jan 19, 2016 at 14:49
  • That's not completely correct. Auto-redistricting would massively change the way US politics works, including reducing ethnic minority representation, or making explicit in a politically untenable way preferences for certain groups. I am pretty certain that even if we had technological savants running legislatures they would not support auto-redistricting, and might oppose it more strongly. This is a status quo bias, but that isn't the same as "relic of the past". Commented Jan 21, 2016 at 16:55

You could write software that would automate redistricting. People have done this. I still don't think that this is the right solution though. Under the current system, politicians choose the districts for themselves (if in the state government) and other politicians (e.g. federal Representatives). Under this system, politicians would choose the software that would choose their districts.

Would it be fairer? Not necessarily. There is no compact geographic redistricting method that will be fair to Democrats in Nebraska or Republicans in Massachusetts.

You can see this in action with the "apolitical" commissions in California, New Jersey, and Arizona. California's districts lean more Democrat than they should by voter registration and statewide vote totals. As do those in Arizona. New Jersey's districts are more Republican than they should be.

Someone generated a set of compact geographic districts in North Carolina where the districts were majority Democrat. Yet North Carolina has a Republican governor and two Republican Senators. The Republican presidential candidate won three of the last four times (and the five times before that; prior to Obama, Carter was the last Democrat to carry North Carolina). So why was that system unfair to Republicans? Why would Republicans vote to switch to a system that is obviously unfair to them?

Given four districts in a state with one large city, should each get a share of the city? Or should the city get its own district and the other three divide up the rest of the state? Both are reasonable. One rule tends to advantage Republicans (cities get their own districts) and another Democrats (divide up the city). In a particular state (e.g. Utah), that advantage may reverse but overall they have a definite lean. Rules can be unfair too.

Under the current system and any compact geographical redistricting scheme, those who are neither Republicans nor Democrats lose out. Those not affiliated with either major party make up more than 40% of the electorate but less than 1% of Congress.

Democrats had overwhelming control of redistricting until 1990, 2000, and 2010 in many states. At the cost of one redistricting cycle, they could have made the changes they recommend to Republicans now. In general, they took the short term advantage of one last gerrymander cycle.

In my opinion, we'd be better served by statewide districts with multiple candidates. These would take politicians out of the process entirely. Voters would be able to select how we wanted to clump. Want to vote for someone local? Go ahead. Want to vote for someone who shares your beliefs? Go ahead. Want to vote for someone of your ethnicity, religion, or gender? Go ahead. You as the voter choose your district. Neither politician nor system can redistrict you.

This would also have the side effect of making voting useful even in areas where one party has an overwhelming advantage. For example, cities routinely vote 90% for Democrats. There's little reason for either Republicans or Democrats to vote in such a district. But with statewide districts, the 10% who don't vote for the local candidate can join with voters elsewhere to elect someone who represents them. Almost everyone would have some representation.

  • I think you have a good argument in here (especially the part about dividing up the city), but as written this answer is way too long and unsourced to make it's point.
    – Bobson
    Commented Jan 19, 2016 at 21:58

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .