A SWEET Cure for Gerrymandering
Achieve nearly proportional representation, destroy incentives to gerrymander, and make every vote matter - all at the small price of one unorthodox feature
Summary
People generally dislike gerrymandering, but they don’t necessarily agree on what they dislike about it or what constitutes gerrymandering. I believe that most of the problems usually attributed to gerrymandering are inherent in the single-representative district voting system used in the United States. That system almost necessarily violates at least two principles intuitively associated with fair elections: proportionality of the outcome to the number of votes cast, and equality of importance of each person’s vote. Nevertheless, solutions based on fundamental changes to the electoral system, such as moving to proportional representation with party lists, are not realistic because they are not compatible with American traditions or legal reality.
I propose a modification to the existing system that would solve most problems with the elections for the U. S. House of Representatives: it would guarantee approximately proportional representation by party, make every vote in a given state count meaningfully, and render partisan redistricting shenanigans futile. The principle governing the modification is that each state’s Congressional representation should be as close to proportional as is feasible, or perhaps within a clearly defined tolerance. The method to accomplish that is to bias the vote share needed for victory (election threshold) in each district as necessary to achieve the prescribed balance. I call this method, and the system based on it, State-Wide Elastic Election Threshold, or SWEET.
The idea that in some districts the candidate with the most votes would lose is unorthodox and I fully expect that most people’s first reaction will be to reject it. However, I argue that my proposal dominates the current electoral system, as well as other proposed solutions, on every reasonable measure of democracy and fairness. I also analyze quantitatively how this proposal would have worked if it had been applied in the 2024 election.
The theory and examples discussed in this article assume a two-party system. That is a good approximation for the current political landscape in the United States. In a future article, I plan to refine the proposal so that it can be applied in more general situations.
Gerrymandering
Recent events have reminded Americans how much they love to hate gerrymandering. Faced with a likely loss of control of the House of Representatives in next year’s election, Republicans decided to gain advantage through redistricting in some Republican-controlled states, most notably Texas, where they expect to gain as much as five Congressional seats. This is possible because Texas is a very big state (38 Congressional districts, second only to California) and its current districts are relatively fair in the sense that its representation tilts Republican only a little more than the vote proportions in the last election; thus, there is considerable room to skew the representation through redistricting. Several other Republican-controlled states are likely to join this effort, but Texas is key to a large shift in partisan balance. (Redistricting is normally done only once following each decennial Census, so doing it again now violates norms and possibly laws, but that is outside the scope of this article.)
In response to Texas’s move, California announced it would pursue a retaliatory counter-measure. As the largest state (52 districts) with a large Democratic majority, California is mathematically in the best position to neutralize the effect of Texas redistricting. The complication was that its law required that redistricting be done by a nonpartisan commission rather than by the legislature. Since that law was enacted as a consequence of a referendum, it could only be changed by a referendum, which was on the ballot this year, one year before the Congressional election, and passed. Several other states have similar restrictions on the power of legislature to draw Congressional districts, which were typically enacted because most people find it distasteful that parties vie for power by gaming the district boundaries. A common slogan, with which it is hard to disagree, says that voters should choose politicians and not vice versa.
I think there is a reasonably broad consensus that “politicians choosing voters” is bad, and any such practice is usually labeled as gerrymandering. But what is gerrymandering? While people may agree on its abstract definition fairly easily, recognizing it in concrete cases is far more controversial. The law is not very helpful in this regard. The only firm principle from the Supreme Court’s jurisprudence is that districts within a state have to be drawn so as to have as equal populations as possible. As I will explain later, this is a low bar and a very naïve interpretation of the “one person, one vote” principle. The other existing rules are that the districts should be contiguous if possible, and that racial minorities have to be reasonably represented (but this last rule is currently being relitigated). Apart from these rather minimal restrictions, states can probably shape their districts any way they want. There have been court cases challenging “partisan gerrymandering”, but every time the courts punted, essentially saying that plaintiffs failed to show clear and workable standards for identifying and rectifying districting wrongs.
As much as I find it obvious that politicians pick voters too often in American practice, and I find it deplorable, I generally agree with the courts that there is no clear legal principle for the judiciary to set the rules in this area and police how they are adhered to. If we are going to have the rules, they will have to come from Congress, which is empowered by the Constitution to set the rules for federal elections. Indeed, Congress has done so in the past; one relevant example for the present discussion is a series of laws enacted through history, culminating in the Uniform Congressional Districts Act of 1967, which mandated that all members of the House of Representatives be elected in single-member districts. In theory, that law could be regarded as a step back for democracy because it codified the (already typical) system that disregards proportionality, but in historical context it was done, at least in part, to thwart the intent of some Southern states to use at-large districts to prevent election of Black representatives. In any case, it is clear that Congress can prescribe the ways states organize federal elections as long as the prescription does not violate the Constitution in some other way.
What Is a Good District?
What principles should guide the design of the Congressional district map? Some that come to mind are: the districts should be geographically compact, they should follow the patterns of clusters of population by demographic or cultural characteristics (and the presumably associated shared interests), they should enable competitive elections to the extent possible, and they should produce outcomes in which a state’s Congressional representation by party (and possibly other characteristics such as race) is reasonably connected to the number of votes cast. In addition, the system should ensure that every individual’s vote matters, so that all people have an incentive to participate in elections. A big problem is that these principles are typically in conflict with each other.
In today’s United States, partisan identity is highly correlated with the position on the urban-rural demographic spectrum. Cities are strongly Democratic, rural areas strongly Republican, and suburbs are somewhere in between (and usually more Democratic the closer they are to the city proper). Additionally, cities are usually more strongly Democratic than rural areas are Republican, so that Democrats, on average, live closer to each other than Republicans and it is easier to “pack” them in a small number of districts. In the 2024 House elections, there were 54 districts where the Republican candidate got more than 70% of the two-party vote and 67 districts where the Democrat won more than 70%. Because of this asymmetry, visually appealing districts of compact shapes or those that have culturally homogeneous (urban/suburban/rural) constituencies are likely to produce election outcomes that give Republicans more seats than their proportion of the vote. For a simple numerical example of how this can occur, consider a hypothetical state named Examplia. It has four districts (see panel 1 in the figure), one of which is a city where 80% vote Democratic, one is rural with a 65% Republican majority, and the remaining two are suburban and lean Republican, say 53% and 54% in an average election. Assuming that the voting (and not just total) populations are of equal size in all districts, Examplia has a 52% to 48% Democratic majority, yet it elects three Republican representatives and only one Democratic in most elections (though that is likely to be reversed in otherwise lopsided Democratic victories such as in the Congressional elections of 2006 or 2018). Winning 75% of the seats with the minority of the vote cannot be considered a fair outcome. As for competitiveness, voters in the urban and rural districts have essentially no influence on the outcome, and the winners in their districts are, for all practical purposes, known before the election begins. Only voters in the two suburban districts can reasonably feel that their votes matter. Although rendering half the voters irrelevant sounds bad, Examplia is much better in that respect than the average real U.S. state: in the 2024 election, the winner got less than 55% of the 2-party vote in only 67 out of 435 (slightly over 15%) Congressional districts. About two-thirds of Americans live in districts that are essentially safe for one party – where victories were more lopsided than 60-40. The situation is even worse in statewide elections (for President and U.S. Senate), where usually only a small number of “swing” states effectively decide the election. Thus, most Americans have no meaningful opportunity to influence any federal elections, and can be expected to feel alienated from the democratic process.
Back to our hypothetical state of Examplia, its four districts could be designed to produce any expected representation from 3-1 Republican to 4-0 Democratic. (It is not possible for the minority party at the state level to win all districts in the state.) It would, for example, be possible to divide it into two safe Democratic districts (combining the city and near suburbs) and two safe Republican, rural-exurbian ones (see panel 2). In one sense, this would seem fair, because the state is close to evenly divided and a 2-2 representation will always be the closest approximation to the vote proportions. On the other hand, that would get rid of the competitive districts and render the elections in Examplia virtually meaningless. At another extreme, it may be possible to make all four districts with roughly the same partisan balance as the whole state. In that case, the expected outcome would favor Democrats (who would, in a typical year, sweep the districts with just a small majority), but all the state’s voters would live in competitive districts that could realistically be won by either party, depending on the national mood and the quality of the candidates. As for the shape of such districts, it may depend on where in the state the big city is; if it is in the center, the four districts might neatly divide the state in four quadrants (panel 3); if, on the other hand, the city is in the corner of the state, the districts would likely have “weird” shapes radiating from the city, resembling the stereotypical image of gerrymandering (panel 4).
It should be clear from this discussion that not all desirable features of electoral districts can be achieved at once. Moreover, “gerrymandering” is in the eye of the beholder, as districts with “normal” shapes can lead to biased representation or predetermined outcomes, and “weird” shaped ones can lead to proportional representation and/or competitive elections. There is no reason to expect that any district map will appear fair to partisans on both sides as well as empower voters. District design is necessarily about choosing priorities, i.e., a political decision. This is true even if they are designed by an “impartial” algorithm or commission, because it will just reflect priorities given to it in its program or charter. For these reasons, I do not believe that a reasonable law “prohibiting partisan gerrymandering” is possible. In fact, I am afraid any such law might result in a worse situation than the one we’ve had. If it prescribed rules for district shapes, such as compactness, it would probably bias the outcomes in favor of Republicans by packing urban Democratic voters in a small number of districts with huge Democratic majorities. If, on the other hand, it mandated that districts be designed in such a way as to make the most likely outcome close to proportional to the votes cast statewide, it would probably feature even fewer competitive districts than we currently have. (Conversely, aiming for as many competitive districts as possible tends to increase deviations from proportionality.) Both cases strike me as bad for democracy.
A Side Note About Voting Power
One fundamental fairness issue illustrated by the hypothetical state of Examplia is the inequality of voting power. Voters in competitive districts decide the election, while those in safe districts are reduced to spectators. This is familiar to anyone who even superficially follows the news from presidential elections, where the campaigns and news coverage are focused entirely on a few “swing states”, while the rest of the country (comprising the vast majority of its population) is taken for granted and ignored. Voters in states like Nevada and Wisconsin have meaningful voting power in presidential elections, while those in California or Tennessee have none. There are ways to quantify voting power in a mathematically rigorous way, and the results show even greater inequality than even a reasonably informed person would guess based on intuition.
It is similar with House elections, just with states replaced with Congressional districts. There is a relatively small number of swing districts, which decide elections, and the rest of the nation just goes through the motions of an election ritual with no hope of influencing the outcome. The differences in voting power are not tenfold or hundredfold; they are more like trillion-fold or even googol-fold. For all practical purposes, the votes outside of the most competitive 10-20% of districts are worth zero.
The Supreme Court has mandated that the principle of “one person, one vote” requires Congressional districts to have equal populations. That rule, in itself, is fair and hardly controversial, but it reflects a painfully minimalist interpretation of “one person, one vote” that places no value on voting power, i.e., the ability to influence elections. Equal-size districts equate votes across districts on an abstract and practically meaningless metric, while doing absolutely nothing for equality of votes in any sense that matters.
The German Solution
Once we realize that the problems we blame on gerrymandering are inherent in the single-member district system, the natural solution appears to be to change the electoral system. For example, many parliamentary systems employ some form of proportional representation, in which parties nominate lists of candidates, voters vote for parties (lists) rather than individuals, and parties win the number of seats proportional to the votes cast; that number of candidates from the top of each party’s list then become members of the legislature. Such a system, however, is completely different from the way elections have been conducted in the U. S. at any point in its history and it is hard to believe that voters could be persuaded to support such a radical change. Besides, the specifics of American federalism would significantly weaken some of the advantages of the proportional system. The U. S. has many states with large differences in population among them, and the Constitution mandates that delegations in the House of Representatives be elected separately by each state. The six least populous states elect only one representative each, and another seven elect two representatives each. Proportional representation is meaningless in those states, and would have little relevance even in somewhat bigger states. This problem could theoretically be overcome by enlarging the House, but only if the number of representatives were made several times greater. While there would be some logic to that (several democracies with about a fifth of the U.S. population have legislatures with more members than the U.S. Congress), it would be yet another radical change American voters would almost certainly not support. Also, as mentioned before, U.S. legislation of election rules has historically trended in the direction of mandating single-member districts. Any reform that would abandon them has to be considered unrealistic.
One major democracy with a proportional system that might seem potentially acceptable to Americans is Germany. German parliament (Bundestag) consists roughly one-half from members elected in single-member districts and one-half from members elected through party lists. Every voter casts two votes, one for the candidate in their district and one for the party list in their state. (Germany is a federation like the U. S., although its states are fewer and more evenly sized.) All candidates who personally win in their districts are elected to the Bundestag, and the party lists fill the rest of the seats in such a way that each party is represented in proportion to the number of votes it received nationally (subject to some restrictions, such as that parties have to win 5% of the party vote or at least 3 individual seats for their lists to qualify).
The German electoral system has many desirable features. It unifies personal representation of local communities with a fair ideological representation proportional to the wishes of the electorate. If we could design our system from scratch (as the entire German system of government was designed after World War II), I would want to copy it. Yet, in the real world, I am afraid it would be too difficult to implement in the U. S., both politically and mechanically.
Even in combination with single-member districts, party lists would be an entirely new feature, unfamiliar to American voters. They would probably be unpopular because, even though American electorate is actually very polarized by party, partisanship carries a negative connotation and both political parties are typically perceived negatively by the majority of people. At a minimum, the House of Representatives would need to roughly double in size to accommodate party list members in addition to those elected in districts, and if we wanted to mimic the legislature-to-population ratio of Germany, the House would need to have about 2,500 members – nearly 6 times its current size. Additionally, a proportionality rule based on the national vote would probably run afoul of the Constitution; instead, it would have to be implemented on a state-by-state basis, limiting its relevance in smaller states. Without unrealistic expansion of the House, this system could also encourage undesirable entrenchment in some smaller states, where the minority party would be essentially guaranteed a seat or two from the party list and could anoint House members with no meaningful accountability to voters.
The New, SWEET Idea
If the German electoral system is so good, but difficult to implement in the U. S., is there a way to replicate its main advantage – combining individual districts with proportional representation – with less departure from the existing U. S. system? There is indeed! Let’s start by identifying the key variables in the German system. Three are obvious: the number of districts won by each party, the proportions of total vote by party, and the number of legislative seats to fill. The first two are outcomes determined by voters (and designers of the districts) and the last one depends on those, according to specified rules, and varies to achieve the proportional representation objective. But there is a fourth variable, which may be hard to notice because it’s “too obvious”: the election threshold (the proportion of the vote required to be elected in a district). We are so accustomed to the rule that to win a district, one has to win more votes than any other candidate, that we think of it as a constant that’s set in stone for all elections. But this threshold is in fact defined differently in different elections. Most U. S. elections (as well as German elections by district) require a mere plurality, some require an absolute majority, perhaps achieved through runoff elections or, more recently, through rank-choice voting, and there are elections where a runoff is required only if no candidate clears a threshold lower than 50% (for example, primary elections in North Carolina require 30% of the vote to avoid a runoff, while primaries for federal and gubernatorial elections in South Dakota require 35%). Having rejected the German choice of treating the legislature size as changeable, the election threshold is the other possible choice of a variable that can be changed according to certain rules to achieve the fairness objective.
Think back to our hypothetical state of Examplia. With the original districting, Republicans won 3 seats with a minority of the vote. A proportionality rule could be imposed that prohibits the representative proportion to differ from the vote proportion by more than one whole seat. The difference between the vote proportion (48%) and representation (75%) is 1.08 seats, which is impermissible. The solution is to flip the winner in one district, and the mechanics of achieving that is to adjust the election threshold in each district by as much as needed to flip one district. In our example, a threshold where Republicans needed 53.01% would achieve this: Republicans would win two districts, with 65% and 54% of the vote, respectively, while Democrats would win the other two, with 80% and 47% of the vote. The last number looks strange, but the statewide election, with votes split 52-48 and representation 50-50, seems fairer. Similarly, under the alternative map with nearly equally balanced districts, Democrats would not be allowed to win all four districts with 52% of the vote (deviating by 1.92 seats from proportional representation. Let’s say the Democratic proportion of the vote in the four districts was 51.3%, 51.7%, 52.3% and 52.7%, respectively; then the threshold of victory would move to 51.31% for Democrats, and Republicans would win one district with 48.7% of the vote. Again, strange-looking outcome in one district, but a fairer one statewide. Since the rule makes the election threshold flexible (or elastic) and is driven by state-wide results, it seems appropriate to call it State-Wide Elastic Election Threshold, or SWEET.
The SWEET Rule, Formalized
Hopefully the working of SWEET is clear by now, but let’s formalize it. Assume there are only two parties (Democratic and Republican), states elect their Congressional delegations in single-member districts, and elections in every district have one candidate from each party. Suppose a state has N Congressional districts, electing one member or seat each. The election threshold in each district is set by default as the majority, i.e., 50% for each party. When the votes are counted, parties get d% and r% of the state-wide vote (vote percentage), where d + r = 100, and majorities in D and R districts, were D + R = N. Therefore, their initial representation percentages are 100*(D/N)% and 100*(R/N)%. Define the tolerance as the maximum allowed difference between the vote percentage and the representation percentage; this tolerance can be expressed as a percentage, number of seats, or a combination of the two (such as “10% or one seat, whichever is greater”). The outcome of the election is then determined as follows:
(a) If the difference between the initial representation and the vote percentages is less than the tolerance, initial representation is also the final representation. In this case, the election is decided in the traditional way.
(b) If the difference between the initial representation and the vote percentages exceeds the tolerance, the election threshold (vote percentage required to win a seat) is changed (biased) in the direction favoring the under-represented party by the minimum amount needed to make the difference less than the tolerance. The final representation (election outcome) is then determined by deciding each district election based on such adjusted election threshold.
Note that the rule specifies that the threshold of victory be adjusted in every district equally, mainly because describing it that way avoids potential ambiguities, but the only districts affected are those that flip due to the application of the rule.
The assumption of two parties and exactly two candidates is a simplifying one. It is a decent approximation of the recent U. S. elections, but it is not an exact representation of elections in many states, nor does it cover all possibilities for future elections. I think simplicity is important at the conceptual level, so this assumption runs throughout this article. I plan to extend the rule and the analysis to more general situations in future work.
Tolerance, the Key Parameter
In our hypothetical example, tolerance was defined as one seat, and it could have equivalently been defined as 25% (1 seat out of 4). The idea, however, is not to have each state define its tolerance, but to make it uniform nationally. For this, a combination of seats and percentages can be useful. It may be too restrictive for largest states, such as California (52 districts) and Texas (38 districts) to keep representation within a single seat from vote proportions. It may even be argued that such strict adherence to proportionality is not desirable, as one of the features of electing representatives in districts is that it tends to augment the win margins, making it easier for the winning party to implement its agenda. A tolerance of, say, 10%, would enable Texas representation to deviate by up to 3.8 seats from the statewide vote share, and Californian representation by 5.2 seats. On the other hand, for most states, 10% tolerance would be too restrictive, as it is equivalent to less than one seat in states with 9 or fewer districts, that is, in all but the 13 most populous states. Our hypothetical purple state of Examplia would be permanently required to elect 2-2 Congressional delegations, making it electorally irrelevant from the national perspective. For this reason, my preferred tolerance specification is “the greater of X% or one seat”.
It should be clear that states with one or two seats are unaffected by SWEET. Obviously, the majority winner must win a state’s single, at-large district, so the six such states have nothing to consider. Also, in a state with two districts, the difference between initial (traditional) representation and vote percentages can never amount to more than one seat (or 50%), as it is not possible to win both districts with a minority of the vote. The existence of two-district states (there are currently seven) provides maybe the clearest case for one seat to be the smallest tolerance ever applied: a smaller tolerance could lock such states to a certain allocation of seats by party for a long time. A tolerance of 25% or less would, for all practical purposes, amount to requiring every one of those states to always send one Democrat and one Republican to the House.
For states with 3 or more districts, the SWEET rule would be relevant. In the next section, I will discuss how actual 2024 U. S. Congressional elections would be affected by this rule, but before getting there, I need to address deviations from the two-party simplifying assumption. The role of third parties and independent candidates was quite insignificant in the 2024 House elections: Democrats and Republicans accounted for over 97% of the total vote. Ignoring third parties and using the percent of two-party vote is therefore a good approximation. However, not all seats were contested by both parties. In 20 districts, there was no Democrat on the ballot, and in 17 there was no Republican. Some of these situations arose in California due to its “jungle-primary” system, where the top two vote winners advance to the general election regardless of party, allowing elections between candidates of the same party. In 2024, one district pitted two Republicans and four districts pitted two Democrats against each other. Elsewhere, some lopsided districts simply didn’t attract a candidate from the sure-to-lose party. This complicates the analysis of statewide votes, and the problem is even worse in some states, such as Florida and Oklahoma, which don’t publish vote tallies in uncontested districts at all.
To make the analysis as relevant as possible, I did it two ways: first, using two-party votes as published (and omitting the two districts without published numbers altogether), and second, adjusting for uncontested (or same-party finalist) districts by replacing vote tallies in those districts (including the two unpublished ones) with those from the Presidential election. Such adjusted tallies are the best proxy we have in the data for the votes that would have been cast had every district featured candidates from both parties in the general election, which would be expected to be the case if this voting method were implemented. Fortunately, the differences between the two methods are small, so I will typically present the adjusted analysis and note the differences in the few cases where they are significant.
My source of data on House elections is Statistics of the Presidential and Congressional Election from Offical Sources for the Election of November 5, 2024”, compiled by the Office of the Clerk of the U. S. House of Representatives, while the source for Presidential Election results by district is “2024 Results for Districts used in 2024” from The Downballot Ultimate Data Guide, https://www.the-downballot.com/p/data.
What Difference Would It Make?
I applied SWEET to the 2024 House election (adjusted as described above) using tolerances of the form “the greater of X% or one seat”, where X is 20, 15, 10, and 5, as well as a straight one-seat tolerance. All cases resulted in small Republican gains, ranging from 4 seats (with 15% tolerance) to 8 seats (with a straight one-seat tolerance). As will be shown below, California accounts for far more deviation from proportionality than any other state (due to a combination of its size, partisan imbalance, and population distribution), so it’s worth mentioning that outside of California, only the 20% tolerance resulted in a Republican gain (of 3 seats), 15% tolerance resulted in no net change, while smaller tolerances produced Democratic gains ranging from 1 seat (with 10% tolerance) to 3 seats (with 5% and straight one-seat tolerances).
It isn’t too surprising that SWEET, which imposes proportionality, would have helped Republicans in the 2024 election. Democrats slightly outperformed, representation-wise, in the House election, winning 49.4% of seats with 48.6% of the two-party vote. (Here I used the actual votes as tallied; with adjustments, Democratic vote was 48.9%.) The difference is equivalent to slightly more than 3 House seats (2 on the adjusted basis). People who know me know that I am a Democrat and might wonder why I would propose a system that favors Republicans. The answer is that SWEET doesn’t favor either party ex ante; its explicit purpose is to maintain a strong relation between Congressional representation and vote proportions. The effect of its application will vary from election to election based on circumstances including details of district maps and voting results, but should generally be fair (according to the principles discussed above). If all the redistricting plans discussed – or rather, threatened – in 2025 are implemented, Republicans would gain some districting advantage, in which case SWEET, designed to blunt gerrymandering, might favor Democrats slightly in the 2026 election.
The summary results of the analysis of effects on SWEET on the 2024 House election (with Presidential election proxies in uncontested districts) are presented in the following table:
One interesting result is that the strictest (one-seat) tolerance would not have achieved the best proportionality of representation nationally; to the contrary, it would have deviated from proportionality the most. The best proportionality would have been achieved using the tolerance equal to greater of 15% or one seat (which would flip half as many districts as the strictest case), and only that one would have improved on the proportionality without SWEET (where the difference between R seats and R vote was –0.51%). Of course, such results would differ from election to election. The 2024 election was somewhat unusual for how little the representation deviated from proportionality nationally. Anyway, this result is important to keep in mind because, if SWEET ever gets implemented, legislators may want to strike a balance between requiring proportional representation and keeping the number of flipped districts low. Understanding that this is not necessarily a trade-off, from the net national representation perspective, may lead to an implementation that achieves good results while upsetting fewer people.
For detailed analysis of effects by state, see the Appendix.
The Challenge and the Response – The Many Advantages of SWEET
The obvious challenge for adopting and implementing this modest modification to the electoral system is the novel feature of biased election thresholds and the intentional result of some candidates winning their districts with a minority of the vote. While exactly equivalent, it sounds even worse that some candidates would lose with a majority of the vote. Why should they and their voters not be upset? Why should we allow such things to happen?
Let me turn this question around: the current electoral system has plenty of bad features; why do we allow those to happen? Why do we allow representation to deviate widely from vote proportions? Why do we allow politicians to draw districts to isolate them from the will of the voters or to increase their party’s power? Why do we tolerate a system in which five-sixths of the voters live in uncompetitive districts and their votes, for all practical purposes, don’t count? I am not asking why the voters are not upset about these features, because there is plenty of evidence that they are. Our electoral system is poorly designed and produces obstacles to democracy that do not appeal to voters and deter some people from voting altogether. Diagnosis is simple: the system is ill and needs a cure. The only question is, what cure?
All cures commonly discussed have significant drawbacks. Radical solutions involving proportional representation may sound appealing in the abstract (in a recent poll, about half the voters expressed support for proportional representation), but would certainly lose a lot of their appeal once voters in larger states understood that the means to accomplish this would be state-wide party lists and loss of representatives’ accountability for local interests. Furthermore, in many smaller states, proportional representation would make election outcomes extremely unresponsive to the changes in the electorate’s mood, possibly even more so than in the current system with a large number of uncompetitive districts. On the other hand, solutions involving legal restrictions on partisan gerrymandering are almost certainly not feasible due to impossibility of reaching a consensus on the precise definition of gerrymandering and criteria for its impermissibility. Moreover, even if an enforceable anti-gerrymandering rule were enacted, it would not solve the problem of most people’s votes not meaningfully counting, as most voters would still live in uncompetitive districts. Clearly, all proposed cures are either ineffective or have prohibitive side effects.
SWEET, admittedly, has the one side effect of having to accept that some candidates who receive more votes than their direct opponents will lose. But this one “weird” feature is the price that buys numerous benefits. It achieves reasonable proportionality of representation, while allowing Congress to specify, uniformly for the whole nation, what is “reasonable” in this context. It removes (or at least significantly weakens) incentives for state legislatures to rig districts for partisan advantage. And finally, what is perhaps the most valuable in the long run, it makes every vote count meaningfully: even if you live in a non-competitive district, your vote counts in the state totals and may move the election threshold and change the outcome in the state’s competitive districts. No voter would be irrelevant, unlike in the current system, where the vast majority are.
Other solutions that preserve single-member districts are incapable of making every vote count. The only alternative way of achieving the same goal I could think of would be allowing people to vote anywhere they choose within their state. Theoretically, that would allow every voter to transfer their vote to the district where it is the most valuable and would achieve an efficient allocation of voting power throughout the state. Implementation of such a system, however, would be fraught with difficulties. The number of voters by precinct would be difficult to predict and some precincts may be overwhelmed by too many voters for the available ballots and staffing. It would also take a lot more resources to verify that each voter voted only once, as all voter lists in the entire state would have to be cross-referenced. It would likely give rise to new partisan games, as parties would try to organize and transport their voters to strategically vote in targeted districts. Besides causing such chaos, it could also be subject to a legal challenge that it sabotages the equal population of districts. None of these problems would occur with SWEET.
As for the weirdness of winning with a minority of the vote, we should keep in mind that some version of it happens all the time. Most U. S. elections do not require clearing 50%, so candidates often win even though the majority of voters voted for somebody else, i.e., against them. By my count, 13 representatives were elected in such a way in the 2024 election; that may seem like a small number until one remembers that the number of competitive districts (those decided by a margin smaller than 10%) is only 5 times as large. Some states (currently only those with significant moose population) have implemented ranked choice voting to prevent this problem from occurring. I am a fan of ranked choice, but it has encountered initial resistance wherever it was proposed. It produces the desirable result (electing the candidate who beats all the alternatives if compared one-on-one), but it is more complicated than traditional voting and sometimes the winning candidate is not the one who got the most first-choice votes. The fact that ranked choice has been gaining acceptance and has been implemented in some states as well as in some local elections (such as the New York City mayoral primaries) is evidence that voters are willing and able to overcome initial resistance to “weird” voting features if they bring tangible benefits.
The objective advantages of SWEET are undeniable, and I think people can be won over to support it in spite of the expected initial resistance. Maybe its biggest problem in the long run is that, by design, it typically rankles the majority. Because single-member districts tend to amplify electoral majorities and also because political majorities in states typically draw the districts, the initial representation (before applying SWEET) is almost always skewed in majority party’s favor, so that SWEET, if it applies, goes against the majority party. (Indeed, this would have been the case everywhere in the 2024 election, although in other elections there are exceptions; for example, in the 2018 election in Wisconsin, Democrats got 54% of the two-party vote, but 5 out of the 8 elected representatives were Republicans.) This means that, after every election, the majority of the voters in affected states will have reasons to be unhappy with the way SWEET affected the election. This creates a potential problem of instability or even democratic legitimacy, where the law is continually opposed by the majority. Fortunately, several factors would mitigate this problem. SWEET is intended to be implemented uniformly nationwide by Congress, not by individual states; it only makes sense as a national rule. Therefore, although it would annoy Democrats in blue states and Republicans in red states, those annoyances would largely balance out, and the reasonably interpreted message to Congress would be that about equally many people in both parties see the rule as unfair to them, so that on the net the rule is fair. As for stability, in most elections, SWEET would not change which party wins control of Congress; but in those few where it would, the party in control would be the one that it helped in the latest election, so it would have no reason to want to change it. Finally, to the extent that voters who constitute the majority in a state are unhappy that the state’s election thresholds were biased against them, they should direct their complaints to the district-drawing authorities in the state and pressure them to design districts that would get closer to proportionality without application of SWEET (although, as discussed, that would not be possible in every state). In other words, not only would SWEET weaken incentives for gerrymandering, but it would also recruit local political majorities to oppose their own party’s remaining gerrymandering efforts.
Conclusion
SWEET is the sweet spot of electoral reform that would solve most of the problems caused by single-member-district electoral system without abandoning that system and losing its few existing advantages. While mechanically designed to achieve approximately proportional representation, it generates even greater benefits by making every vote meaningful and curbing partisan gerrymandering. It achieves this at the price of requiring the outcome of elections in a few districts to be decided in a way that challenges tradition and intuition, but its unique alignment with the sentiments of most voters in several important areas should enable it to win popular support. Furthermore, as far as I can tell, it is fully compatible with constitutional constraints and its implementation would not cause any practical difficulties.
The discussion and analysis in this article is based on a simplifying assumption that there are only two political parties. This is a good approximation of the current political reality and a good basis for educating the public about SWEET. Actual implementation would need to account for more parties and ensure compatibility with other desirable electoral reforms, such as ranked choice voting. This is important because one of the possible results of SWEET could be viability of more than two political parties. I plan to extend the formal statement of the SWEET rule and its analysis to the more general case in future work.
Appendix: Detailed Effects by State
Which States Deviate the Most from Proportionality?
Ignoring one- and two-district states, to which SWEET wouldn’t apply, the largest percent differences between vote and seat proportions are found in New Mexico (3-0 D with 55% of the vote), Iowa (4-0 R with 57%), Connecticut (5-0 D with 59%), Nebraska (3-0 R with 64%), Massachusetts (9-0 D with 64%), Utah (4-0 R with 66%), Arkansas (4-0 R with 68%) and Oklahoma (5-0 R with 69%). Using any of the five levels of tolerance tested, most of these states would have been awarded one seat for the minority party, except CT, where Republicans would have gained two seats, and MA, where Republicans would gain between two and three seats. To accomplish this, the vote percentages needed for Republicans to win districts would have been 47.92% in NM, 50.10% in IA, 41.98% in CT, 50.93% in NE, 62.89% in UT, 58.94% in AR, and 60.70% in OK. In MA, the threshold would be 43.04% for a two-seat switch and 40.30% for a three-seat switch.
Massachusetts is one of the states where adjustments based on the presidential election were especially large because, in the actual House election, Republicans fielded candidates in only 2 of its districts, winning only 9% of the total vote and 11% of the two-party vote. (Due to the absence of Republicans, other parties were relatively visible in some districts, so that Democrats won 68% of the total vote cast, but 89% of the two-party vote.) It is also a showcase for a state that would almost certainly have more contested elections with a system that valued proportionality. Deviation from proportionality in Massachusetts is not due to gerrymandering; rather, it is impossible to draw even a single Republican-majority contiguous district because Democrats are a majority throughout the state. For example, Democratic presidential candidates have won every Massachusetts county in every election since 1992. Therefore, no anti-gerrymandering rules could fix the lack of Congressional representation of Massachusetts Republicans
In terms of the number of seats, by far the greatest contributor to deviation from proportionality is California, where Democrats won 43 out of 52 seats (83.7%) with 60.7% of the vote (or 60.2% adjusted). This amounts to a difference of 11.5 (or 11.7) seats. Such a large number is primarily a function of California’s size (only 3 other states have half or more of its districts and 5 more have a quarter), but its percentage deviation is also rather large, putting it roughly in the middle among the states. This may be surprising considering that California’s districts were drawn by a nonpartisan commission, as required by law. Indeed, the shape of California’s districts generally does not visually appear controversial. But California is a strongly Democratic state and, while the partisan lean is not as uniform as in Massachusetts, the vast majority of Californians live in an area with just a few pockets of Republican majority. If one tried to draw districts in California that would result in a proportional representation (31-32 Democrats and 20-21 Republicans), it would look extremely gerrymandered, if it would be possible at all. Furthermore, California has a fairly large number of competitive districts. It has 9 of the nation’s 67 districts that were won by less than a 10% margin, which slightly more than California’s share of all Congressional districts, despite it being a state with an overall 20% partisan imbalance. Maximizing the number of competitive districts is a desirable feature of district maps, but it also tends to favor the statewide majority party, as we have seen on the example of a stylized hypothetical 4-district state. A final contribution to California’s deviation from proportionality is that Democrats were lucky to win two extremely tight elections, by margins of 187 votes in district 13 and 653 votes in district 45. Thus, shifting two seats, which would satisfy the 20% tolerance, would only require moving the Republican threshold of victory to 49.90%. 15% tolerance would require moving it to 48.56% to shift 4 seats, 10% tolerance would require a 47.42% threshold to shift 7 seats, and the strictest requirement of no more than a seat difference would require setting the threshold at 42.38% to switch 11 seats.
After California, the state with the most seats away from proportional representation is Illinois, where Democrats won 14 of its 17 districts (82.35%) with 56.30% of the vote (52.92% without adjustment because there were two districts with no Democratic candidate). To be within a seat of proportionality, Illinois’s House delegation would need to be 10 D to 7 R, and to switch 4 seats to Republicans, their threshold of victory would need to be 44.40%. A threshold of 44.89% would switch 3 seats, satisfying 10% tolerance, and one of 45.57% would switch 2 seats, satisfying 15% and 20% tolerance.
Besides the already mentioned Massachusetts, two states could be required to switch up to 3 seats. One of them is New York, where Democrats won 19 of the 26 seats (73.08%) with 58.75% of the vote. Balancing New York is Florida, where Republicans won 20 of the 28 districts (71.42%) with 57.15% of the vote. Both states’ elections would be unaffected under 20% and 15% tolerance levels, but under a strict one-seat tolerance it would be necessary to switch 3 seats – to Democrats in Florida, requiring 56.53% for a Republican victory, and to Republicans in New York, setting the threshold at 46.49%. Switching 2 seats would satisfy 10% tolerance, requiring a threshold at 54.83% in Florida and 47.09% in New York.
States not already mentioned that would be required to switch up to two seats would be Texas, North Carolina and Tennessee (from R to D) as well as New Jersey and Washington (from D to R). Pennsylvania, Ohio, Georgia, Arizona, Indiana, Missouri, Wisconsin and South Carolina would need to switch up to one seat from R to D, while Maryland and Oregon would need to switch one seat from D to R.
Unaffected States
Several states have very nearly proportional Congressional representation under current rules. Minnesota, Virginia, Michigan, Louisiana, Colorado and Alabama deviate by less than 7% from proportionality and would not be required to switch any seats under any of the tolerances tested. Additionally, Mississippi, Kansas, Kentucky and Nevada deviate by larger percentages, but still less than by one full seat, so they also wouldn’t be required to switch any seats. That makes 10 states that could in principle be affected by SWEET, but are not, in addition to the 13 states that could not be affected because they have only one or two districts. It is notable that in several of these states (AL, LA, MS) the minority-majority districts currently required by law are sufficient to bring the state representation into the proportional zone.