Executive Summary

• State legislatures are essential to the American republic. While much attention is given to Congress, state governments make most of the decisions that impact day-to-day life. While congressional districts typically cover over 700,000 people and leave many voters locked out of representation, state legislative districts are smaller and theoretically give voters another chance to be represented.
• However, like Congress, most state legislators are elected from single-member, winner-take-all districts. Single-member districts are vulnerable to disproportionate representation – when certain groups (e.g. partisan and racial groups) receive fewer seats than their share of the vote would suggest.
• In particular, single-member districts often award the majority party a “winner’s bonus,” and deprive racial minorities of the “power to elect” candidates of their choice when they are geographically dispersed.
• This report measures partisan and racial disproportionality in state legislatures, and ranks the states in terms of “worst representation.” Representation is measured by “power to elect,” which measures the ability of a racial or partisan group to fill a seat without relying on outgroup voters.
• Proportional ranked choice voting (RCV) is a replacement for single-member districts that promotes better representation and projects voting rights. Proportional RCV makes it easier for minority racial and partisan groups to gain fair representation by lowering the “threshold to elect,” and by allowing more than one partisan and/or racial group to gain representation within a district. This report gives examples of how proportional RCV would improve representation for racial and political minorities in five states – Michigan, New Jersey, North Carolina, Oklahoma, and Oregon.

Introduction

Proportionality, the idea that votes should directly translate into seats, is not guaranteed in most American elections. That’s because most legislative elections in the U.S. use single-member districts with a single-choice, winner-take-all election. 

In a winner-take-all election, the candidate (or party) with the most votes represents the entire district. Therefore, single-member districts don’t differentiate between a district won by 51% and a district won by 99%.  The choices of political minority voters in a winner-take-all election can be said to be “wasted,” as they have no effect on the winner.

If a certain political or racial group makes up a significant portion of a district, but not quite 50%, that group will not be represented in the district. If this happens in enough districts, it results in disproportionate representation – a gap between a political group’s share of votes and its ability to elect representatives. For example, about 39% of Bay Staters vote Republican, but only 8% of seats in the Massachusetts State Senate are safely Republican. In North Carolina, Black voters make up about 21% of the voting age population, but only 3% of districts in the North Carolina General Assembly are majority-Black.

When population patterns allow, federal law requires single-member districts to be drawn in a way that maximizes proportional representation for minority racial groups. However, drawing majority-minority single-member districts can be difficult, or impossible, when voters of a racial minority are geographically spread out. If a minority group is not a majority of voters anywhere, then a district cannot be drawn which gives them the ability to elect a seat. Securing fair representation with single-member districts can be difficult when a racial group is large in number, but not geographically compact. 

When it comes to partisan representation, those drawing district maps often go out of their way to skew results in favor of their preferred party – a practice known as “partisan gerrymandering.” The governing party may draw the lines in a way that prevents the opposition party from reaching 50% in a fair number of districts, insulating itself from electoral loss. Particularly notable in this regard are states like Kentucky and Wisconsin, where one party has drawn itself into an overwhelming majority of districts, meaning they are likely to keep control of the legislature even if a majority of voters select the other party’s candidates.

This report will show the extent of disproportionate representation in state legislatures, and compare each state in terms of political and racial representation. This report will also present multi-member districts and proportional ranked choice voting (proportional RCV) as a vehicle for fair representation. By applying the “Fair Representation Act” model at the state level, more voters would have power to elect in their districts. To illustrate this idea, we drew multi-member redistricting plans for five state legislatures to demonstrate proportional RCV’s benefits.

Methodology

To quantify partisan and racial representation, we collected data from shapefiles on state legislative districting plans in all 50 states – publicly available on Dave’s Redistricting App. Each jurisdiction’s political composition was estimated by a composite of the 2016 and 2020 presidential elections, as well as any U.S. Senate elections, gubernatorial elections, and attorney general elections. Voting age population (VAP) data was sourced from the 2020 Census.

To measure racial representation in each state legislative chamber, we calculated the share of districts in which each racial group comprised greater than 50% of the VAP, and compared that to the respective group’s share of the statewide VAP. The difference between the two represents a “skew,” or the degree to which representation is disproportionate. The closer a state’s skew is to zero, the closer that state is to proportional representation.

For winner-take-all districts, we used 50% as a power to elect threshold. Compare this to a district where a certain racial group makes up 45% of the VAP; theoretically, a candidate could win an election without a single vote from that racial group. However, there are a few limitations to this approach. We did not attempt to factor in varying levels of racially polarized voting or White support for minority candidates. Therefore, the power to elect threshold doesn’t represent an actual opportunity district based on racially polarized voting patterns, but instead means a racial group can elect a candidate of its choice if it votes cohesively, or that any winning candidate must draw at least some votes from that group. 

Therefore, power to elect does not indicate whether a member of the racial group actually holds the seat. This way, our data is applicable across election cycles. This is fairly evident in North Carolina’s result. Black voters only surpass 50% of the VAP – and therefore have the power to elect – in four of the North Carolina General Assembly’s 120 districts. Yet the assembly has 26 Black members, elected mostly from plurality-Black or narrowly plurality-White districts (This is partly because Democratic districts where Black voters comprise the majority of the primary electorate tend to elect Black representatives). A change in methodology¹ (counting plurality, but not majority, Black districts as Black power to elect seats) would have roughly halved the state’s measured skew, but the skew would still be significant.

To measure partisan representation, we directly assigned districts as either “Democratic” or “Republican” based on the 2016-2020 political composite. Note this is a simplification; we considered a district with a 51% Democratic majority a “Democratic seat,” when pundits would consider it a “toss-up seat.” Since toss-up seats are most often won by either a Republican or a Democrat, assigning “toss-up” districts to one of the major parties creates a more accurate picture of a chamber’s partisan composition. For this reason, our partisan categorizations represent power to elect rather than which parties actually hold seats. This way, our data is applicable across election cycles. 

We then compared the shares of “Democratic” and “Republican” districts to the shares of “Democratic” and “Republican” voters statewide. Again, the difference between the two represents a “skew,” or a level of disproportionate representation. To achieve proportional representation, the ideal skew would be close to zero.

Next, we compared all of the state legislative chambers in terms of their “skews” for partisan and racial representation. The results are compiled below.

Results

Our results are compiled in four sortable tables. The tables show:

• Skew of partisan representation in state lower houses
• Skew of partisan representation in state senates
• Over- and under- representation of White voters in state legislatures
• Over- and under- representation of individual racial groups in state legislatures

The “skew,” or percentage-point gap between each political and racial group’s voting age population and its power to elect, is measured for each state legislative chamber. Be aware that rankings for the lower house are out of 49, not 50 (Nebraska has no lower house).

Partisan skew

First, we reflect on partisan representation. Large partisan skew was highly correlated with large partisan statewide majorities: Heavily blue or heavily red states had the largest skews. This is causally explained by the “winner’s bonus” effect of single-member districts. Essentially, if a party wins an election by greater than 50%, it is expected that it will control a greater share of representation than its share of the vote. As Wang (2016) explains it, “single-member districts usually generate outcomes in which a party’s share of seats tends to exceed its proportion of popular support.” The reasoning is fairly intuitive; if a voting bloc is a minority of voters in a given state, and is spread out evenly throughout the state, it is unlikely to have a majority in any particular district. For instance, if Group A earns 51% of the vote in every district and Group B earns 49% in every district, then Group A wins 100% of seats and Group B wins 0%.

In reality, voting blocs are rarely spread out evenly across a whole state, so minority groups often do make up a majority in some districts. Yet even in these cases, minority groups usually don’t make up a majority in enough districts to earn representation equivalent to their population share. Thus, although the extent of the winner’s bonus varies based on local political geography, the winner’s bonus pattern has been observed across time in state, federal, and international elections that use single-member districts. It is applicable to both political and racial groups. Therefore, partisan disproportionality is a natural result of single-member districts.

Gerrymandering can exacerbate partisan disproportionality, but isn’t the root cause. To illustrate this, Wisconsin was one of the most gerrymandered states at the time of data collection, but its lower house ranks only 24th worst in terms of partisan representation, and its upper house ranks 17th worst. Regardless of how it occurs, a skew in representation affects democracy. An outsized legislative majority for a given party changes the behavior of the state legislature by shifting the median lawmaker away from the average voter – as shown by Caughey, Taysanovitch, and Warshaw (2017) in the context of gerrymandering.

Partisan skew rankings

Racial skew

Moving from political to racial proportionality, the data reveals that Oregon’s state legislature has the worst under-representation for voters of color of any state. Voters of color have the power to elect just one seat in the lower house (2% of the chamber) and no seats in the upper house, a remarkable occurrence in a state where 25% of voters are people of color.

Representation in Oregon is so skewed because voters of color tend to be geographically spread out across the state. Oregon has several significant communities of color – especially of Black, Native, and Hispanic or Latino voters – but they’re concentrated in different regions of the state and don’t make up over 50% of the vote in those areas. That makes it impossible to draw many majority-minority districts in most cases – even “coalition seats” where voters of several racial groups could vote in a coalition to elect a candidate of their choice.

Oregon’s example demonstrates a strong case for proportional RCV. By utilizing multi-member proportional representation, the threshold for representation is decreased. Even if voters of color do not make up a majority in any district, they can still elect a representative of their choice with a proportional system.  

Oklahoma also provides an opportunity to examine poor racial proportionality. In the Oklahoma State Senate, every district has a majority of White voters. In the lower house, the situation isn’t greatly improved: Only 13 of 101 districts  aren’t majority-White, which is abysmal considering 35% of Oklahoma voters are not White. This pattern is primarily a result of Native American voters (Oklahoma’s largest minority community) being dispersed roughly evenly throughout the state. This results in many areas with large numbers of Native voters, but few places where there are more Native voters than White voters.

Additionally, most Native voters in Oklahoma live in rural counties, while Oklahoma’s other voters of color, like Black and Hispanic voters, tend to live in urban areas. These two factors make “coalition seats” very hard to draw in Oklahoma. In short, even the most benevolent map drawer would find it difficult to provide fair representation to Oklahoma’s voters of color.

In Oklahoma, Oregon, and almost all states, racial proportionality is marginally better in the lower chamber than the upper chamber. This is because in every state legislature, the lower house has more seats – and therefore smaller districts – than the upper house. In general, it is easier to achieve minority representation when districts are smaller, because smaller areas where a minority community makes up a majority can be given a seat. With a larger district size, those minority voters may need to be combined with White voters in order to reach the population requirement.

Smaller districts improve racial proportionality by lowering the raw number of votes needed to elect a representative, but do not change the 50% threshold to elect a presentative (the “power to elect” threshold) that is consistent across winner-take-all elections. The fundamental reason that single-member districts fail to deliver racial proportionality is the 50% threshold; a racial group must be the majority of voters in enough precincts to have the power to elect their representative of choice through majority control of a district. This only happens when racial communities are geographically concentrated; voters of color cannot be given the power to elect a representative in a single-member district if they do not constitute a majority of votes anywhere, regardless of district size.

While a handful of state legislatures do use multi-member districts, they all use winner-take-all voting methods that allow one majority-group to elect every winner – such as Bloc or Post elections. For instance, we could not find a single case in the 2022 elections of a multi-member state legislative district not electing all of its representatives from one party. Because these multi-member districts retain the winner-take-all effect of single-member districts but cover a wider area, they actually worsen disproportionality and have historically been used to thwart fair racial representation.

To view racial skew metrics for every state legislative chamber in the country, check out the visualizations below.

Racial skew rankings

In our analysis, we considered a racial group to have power to elect when that group makes up over 50% of voters in a district. In these districts, a racial group can elect its own preferred candidate if it so chooses, without needing the support of any other group of voters.

Power to elect is an important concept, because it conveys the amount of influence a racial group can exert by simply achieving internal consensus. There are also some districts where no individual racial group constitutes a majority of the voting population. For example, North Carolina’s 40th Senate District is a “coalition seat,” since the population is 39% Black, 35% White, 19% Hispanic or Latino, and 7% Asian and Pacific Islander. Coalition seats provide opportunities for coalition building; candidates may be incentivized to make cross-racial appeals, because they could not get elected with one group’s support alone. In our analysis, coalition districts are considered majority-minority districts.

However, power within a coalition is not the same as sole power to elect. Voters of color are not a monolith, and shouldn’t have to be lumped together to have voting power. For this reason, coalition seats were not considered in our analysis of individual minority groups (i.e. the “minority underrepresentation” chart); we don’t attempt to assign the largest minority group in a coalition seat with control of that seat.

Similarly, individual racial communities are not a monolith, and may not vote cohesively nor have the same political preferences. Though the 50% threshold represents a community’s power to elect a candidate of its choice, on a more basic level, it means a candidate can’t win without at least some of that community’s support. 

Therefore, by not attempting to model levels of cohesion in how racial groups vote, and by not attempting to assign “coalition seats” to any racial group, our measurements in some states can differ from how often a racial group elects a candidate of choice in practice. However, this does not affect our findings for overall White overrepresentation, and only greatly affects a few states with many coalition seats.

One state where counting coalition seats would make a significant difference is Maryland. In the original dataset, Black voters in Maryland have the power to elect nine state senators. Counting districts in which Black voters are a plurality adds five more districts. This change brings the Maryland State Senate’s skew for Black voters from -11.86% to -1.22%, within one seat of perfect proportionality. This change improves the Maryland Senate’s ranking in Black representation relative to others by 34 places, and was the largest change observed. 

Likewise, Black voters make up a plurality in 5% of districts in the Illinois House, and 4% of districts in the Georgia House. Hispanic or Latino voters in New Mexico were the plurality of voters in 14% of districts – which, if counted as “Hispanic” controlled districts, would put Hispanic voters in control of 50% of the seats, close to their 45% of the voting age population. Texas, New York, and California all saw a significantly smaller gap in Hispanic representation in both chambers of the state legislature when coalition seats were counted. See the footnotes for an alternate version of our data which counts plurality racial groups as having power to elect their seats. 

Multi-member districts and proportional RCV: a better way forward for representation

Majority-minority single-member districts have long been the default resolution for voting rights violations caused by winner-take-all bloc voting. However, the remedy’s success depends on cooperative line drawers and continued support from the courts. Additionally, as our analysis shows, single-member districts often fail to deliver representation when minority communities are geographically dispersed.  

In recent years, proportional representation has been discussed more and more as an alternative remedy for voting rights violations in local elections. Proportional ranked choice voting is now outlined as a remedy in four state voting rights acts, including its recent introduction as an option in New York and Connecticut. 

Proportional RCV utilizes multi-member districts. Instead of one candidate winning with a majority of votes, several candidates win with smaller shares. Therefore, the “threshold to elect” is not 50%, but a lower number, depending on how many candidates will be elected. For a three seat election, the threshold is 25% + 1 vote. For a five seat election, the threshold is 16.7% + 1 vote. As a result, the threshold for representation is within reach for more voters of color.

Proportional ranked choice voting can bring relief for minority voting groups, even in jurisdictions where it is impossible to draw a majority-minority single-member district. For example, Albany, California used to use bloc plurality voting to elect its five city councilors. Voters could vote for as many candidates as there were seats up for grabs. This system allowed the majority “bloc” to essentially elect every member of the council. No racial group alone makes up a majority in Albany, but White voters comprise the most at 47%, followed by Asian voters at 30%, and Hispanic or Latino voters at 12%. After voters decided to switch to using proportional RCV, Albany used it for the first time in 2022 to elect two councilors, meaning the threshold to elect was 33%. Voters elected two candidates of color who ran as a slate and represented different areas of the city: John Miki and Robin López.  

Currently, if a government can prove racial underrepresentation cannot be remedied with redrawn single-member districts, that government can often escape litigation. This is no longer the case in states where proportional RCV is a sanctioned remedy. 

Proportional ranked choice voting also remedies another undesirable property of single-member districts: gerrymandering. One reason is that using larger (and fewer) districts means there are simply fewer opportunities to draw malign boundaries. Another reason is that proportional RCV makes it far more difficult for legislators to draw lines for partisan gain. A gerrymander functions by wasting the opposition party’s votes through inefficient victory margins. In a two-party winner-take-all race, a maximum of 49% of the vote can be “wasted” on losing candidates. Compare this to a five-member proportional RCV election, where a maximum of only 16.7% of the vote can count toward losing candidates.

Any election method that shifts away from a “majoritarian” system (like plurality voting or single-winner RCV) and toward proportionality makes gerrymandering far more difficult. For this reason, proportional RCV has been put forward as a solution to gerrymandering by scholars like Michael Li of the Brennan Center. Additionally, proportional RCV can be implemented in one multi-member, at-large district for many municipal councils, eliminating the need for potentially divisive redistricting at all. 

Proportional RCV would almost certainly give Democrats in rural areas and Republicans in urban areas the power to elect representatives of their choice. Not only would this ensure fewer votes are “wasted,” but it also means new voices and interests would be represented in legislatures. 

Demonstrating improved representation with proportional RCV sample maps

Proportional ranked choice voting would improve representation in state legislatures around the country. We have drawn up sample multi-member maps for a handful of states, and projected their election outcomes, in order to compare the results to the states’ existing single-member plans. To draw each multi-member map, we either worked from scratch, based our map off combining existing districts, or modified a computer generated plan (using a method called Markov Chain Monte Carlo, or MCMC). 

Disclaimer: All of our maps are merely samples, and are not necessarily a reflection of what the real maps would look like if proportional RCV is adopted in each state. However, many of these maps were drawn with various limitations in mind, particularly the preservation of existing House district lines, which is evidence that these proportional RCV maps would produce results reflective of proportional RCV in general, not just that one plan.   

Consistent with FairVote’s general guidelines for proportional RCV elections, most districts elect either three or five representatives. In districts with five representatives, any candidate earning at least 17% of the vote wins a seat. In districts with three representatives, any candidate earning over 25% of the vote wins a seat. These thresholds create the opportunity for more groups to have a seat at the table than in single-member districts, where candidates need a majority or near-majority to win a seat. 

Also note that because of how the Census counts biracial people, the sum of racial voting age populations (VAP) adds up to either slightly over or slightly under 100%. This discrepancy is under 0.5% in most states, but in Hawaii it is over 15%. Therefore, we corrected this error in Hawaii as a special case. 

Michigan State Senate

Michigan’s State Senate is currently elected via 38 single-member districts. In our sample map, Michigan’s Senate would be elected via eight multi-member districts, seven of which would elect five members and one of which would elect three members. This sample plan was originally drawn based on a computer-generated map, which was cleaned up and adjusted to match local jurisdictional borders and improve compactness.

Sample map

Better partisan representation

Considering only the major parties, Michigan’s statewide partisanship is 52% Democratic and 48% Republican. Under the current maps, Democrats and Republicans each have an advantage in 19 seats. Because Michigan uses an independent redistricting commission and is so narrowly split between parties, proportional RCV would likely not have a dramatic effect on partisan control.

Better racial representation

However, proportional RCV would have a significant effect on racial representation. Under our sample map, the number of seats with majorities of voters of color doubles, from four to eight. Black voters gain the power to elect three additional representatives. 

Practically, when taking into account racially polarized voting and White support of minority-preferred candidates, the current single-member district map features six opportunity districts for voters of color. Proportional RCV’s eight opportunity districts are still a significant improvement, especially considering that we do not include the equivalent of racial “crossover districts” (i.e. majority-White districts where there is significant White support for minority-preferred candidates) here.

Oklahoma State Senate

Implementing proportional RCV would transform Oklahoma’s State Senate map from 48 single-member districts to 10 multi-member districts. One district would elect three senators, and the rest would elect five. 

The sample map was drawn from scratch, and attempted to split county boundaries a little as possible. The map was drawn to comply with the Voting Rights Act by maximizing opportunity districts for racial minority groups.

Sample map

Better partisan representation

Considering only the major parties, Oklahoma’s state partisanship is 64% Republican and 32% Democratic. Under the current single-member district map, Republican’s dominance in the Senate greatly eclipses their share of the vote. Proportional RCV would preserve the Republican majority, but give Democrats a share of seats more representative of their electoral support.

Better racial representation

With proportional representation, Native voters have the power to elect five representatives, compared to zero under the current map. Native voters are not a majority in virtually any precinct in Oklahoma; it is extremely difficult to draw a single-member district for the Oklahoma Senate that contains a majority of Native voters. With proportional RCV, the large yet dispersed population of Native voters gains the power to elect representatives. The number of racial minority controlled or coalition seats is increased from five in the current map to 14 with proportional ranked choice voting.

Oregon State House

Oregon’s State House currently elects its 39 Democrats and 21 Republicans through 60 single-member districts. Oregon’s districts are nested, meaning each State Senate district is a combination of two State House districts. Reflecting this custom, and also to demonstrate that the representational benefits of proportional RCV are not a consequence of precisely drawn maps, our sample map combines existing Oregon State House districts into 16 multi-member districts. The sample map was entirely hand drawn, and includes 10 three-member districts and 6 five-member districts. Note that there is some deviation between our sample map and State House lines – largely where State House lines split precincts in Districts 1 and 16 – but in one case to maintain contiguity.

Sample map

Better partisan representation

Considering only the major parties, Oregon’s statewide partisanship is 57% Democratic and 43% Republican. Proportional ranked choice voting would preserve the Democratic majority in the State House, and award a more fair share of seats to Republicans.

Better racial representation

Under our proportional RCV sample map, Hispanic or Latino voters in Oregon would have the power to elect one additional seat, for a total of two seats. Additionally, Oregon would have 10 coalition seats where no single racial group can elect a winner. As mentioned in the full report, voters of color are more underrepresented in the Oregon State House than any other state legislative body in the nation. For geographic reasons discussed previously, this is difficult or impossible to remedy with single-member districts. Proportional ranked choice voting is a viable alternative that gives more power to communities of color. 

North Carolina

The North Carolina State House is currently elected via 120 single-member districts. Under our sample map, the State House would be elected via 24 five-member districts. The map was constructed purely by combining North Carolina’s existing State House districts. An MCMC program was run to find a combination of districts that would maximize representation for minority racial groups. This map was chosen out of several similar options because of its compactness.

Sample map

Better partisan representation

Considering only the major parties, North Carolina’s statewide partisanship is 50.3% Republican and 49.7% Democratic, yet the current maps heavily under-represent Democrats due to a partisan gerrymander. Proportional ranked choice voting would bring the State House closer to partisan parity.

Better racial representation

With proportional RCV, the North Carolina House would see a dramatic increase in the number of senators whom Black voters would have the power to elect. Black voters would not have to join in coalition with other groups to elect their preferred candidates.

New Jersey: State House and State Senate

New Jersey’s state legislature is composed of a 40-member Senate and an 80-member General Assembly. However, New Jersey has only one set of state legislative districts (i.e. it uses the same map for both chambers); each district elects one senator, and two members of the assembly using block voting. This idiosyncrasy is preserved in our sample proportional RCV plan. For the Senate, there are 12 three-member districts and 1 four-member district. For the General Assembly, there are 12 six-member districts and 1 eight-member district. This is a departure from FairVote’s usual recommendation of 3 and 5 member districts for proportional RCV, but is necessary to preserve the nesting of districts. 

The sample map was based on a computer generated (MCMC) map fitted so as not to split up neighborhoods, to the best of our ability.

Sample map

Better partisan representation

Considering only the major parties, New Jersey’s statewide partisanship is 56% Democratic and 44% Republican. Democrats are overrepresented in both houses of the legislature. Proportional ranked choice voting would preserve the Democratic majority, while awarding Republicans a fair number of seats.

Better racial representation

The New Jersey General Assembly would be composed of six- and eight- member districts. In the districts that elect six representatives, any group that makes up 14.3% of voters can elect a candidate of its choice. In the district that elects eight representatives, this threshold would be 11%. This plan would give Hispanic or Latino voters the power to elect five additional representatives, Black voters four additional representatives, and Asian voters four additional representatives. This is a particularly significant improvement for Asian voters, considering they don’t currently comprise a majority of any district. The projections for the State Senate are similar, if more modest.

Conclusion

The racial and political disproportionality in state legislatures outlined in this report incur tangible cost to the quality and representativeness of our elections. This is a foreseeable effect of the winner-take-all districts that state legislatures use universally as their method of election. However, moving to proportionality does not require a total overhaul of American elections. By employing ranked choice voting in multi-winner elections, state legislatures can deliver proportionality with the same ranked choice ballot that 13 million Americans have already voted with.

Footnotes

1. Running our data with a counterfactual methodology that assigned plurality controlled seats to a specific racial group did not significantly change the relative rankings of states. We defend our decision to not equate these districts and true majority-minority districts, because minority representation in a “coalition” district requires votes from outgroup voters. This, and the fact that analysis of racially polarized voting would involve many subjective decisions on the part of the researcher, led us to use the simple majority approach and count coalition districts separately.