Legislatures Should Be Representative

And ours is not

Why Legislatures Should Be Representative

One might think that the role of representatives in a democracy is simply to record which policies command majority support—that is, to act in place of a public too busy to cast its own vote on every bill. This plebiscitary model does not, however, accurately describe lawmaking. On one level, legislatures are arenas for debate and moral suasion, and we rightly hope that legislators are sometimes swayed by what they hear. More fundamentally, however, legislation is not passed one-by-one in isolation, but as part of a composite process in which votes are exchanged across issues to secure a governing majority. A policy whose large benefits are concentrated in a minority and whose modest costs are spread across the majority would always be rejected in a self-interested plebiscite; yet it may prevail in an equally self-interested legislature, where that minority can gain allies by offering its support on unrelated bills. This phenomenon of quid pro quo is called “logrolling,” and is key to understanding how deliberative democracy considers not just the number of people it affects (“extent”) but the degree (“intensity”).

• Hypothetical 1: Suppose a $100 million dam would yield $2 billion in benefits from cheaper electricity and flood control to 10% of the population (1 million people). We might predict that the other 90%, paying a cost but seeing no benefit, would vote against it—leaving society $1.9 billion poorer. This is a supposed flaw of democracy: that extent always beats intensity (“two wolves and a lamb voting on lunch”). Yet, this is not what we see in practice: despite Interstate 5 only running through 6% of states, the other 94% do not use their majority kill it. Instead, it gets bundled with every other interstate—which 100% of states have. (Even Alaska and Hawaii have “Interstate Highways,” despite those highways not going inter-state.) Thus, while the plebiscitary model predicts that the dam with high ROI would fail since it only serves 10% of the country, the composite model correctly anticipates the possiblity of there being multiple dams.

  • Seldom, however, are projects actually as simply equivalent as “everyone gets a highway” or “everyone gets a dam.” Large dams are simply less useful along the East Coast than in the Mountain West, so what matters for logrolling is not identical projects but equivalent returns. An $800 million infrastructure package might generate $16 billion in benefits by pairing forty harbor upgrades in the East with four major dams in the West. Such bundling lets valuable but parochial projects—each too narrow to pass alone—attract a viable coalition.

• Society-enriching cuts (whether to spending, taxes, or regulation) work the same way as the society-enriching spending described above. If the country as a whole could benefit from a lower budget but there is a majority behind every specific spending item, legislative leaders can choose a number to cut and then logroll to the least painful combination of cuts. The final package would be able to win a majority even if every specific cut only has minority support.

• This process of mutual accommodation and negotiation works better the more closely the legislature resembles the underlying population so that the legislature appropriately weighs trade-offs.

  • Hypothetical 2: Consider another hypothetical dam of similar design to the one discussed above. This one, while costing the same, would only provide its benefits to an area with 1/40th the size and population. We would hope that the legislature would decide not to fund it—$100 million in costs for $50 million in benefits is not society-enriching. Yet, since Dam 2 superficially has the same profile of supporters as Dam 1 (a small minority incentivized to use its representation to aggressively logroll for it), it isn’t immediately obvious that the reduced number of beneficiaries will actually prove decisive.

    • It turns out that, when a legislature accurately mirrors the population, extent and intensity automatically balance. Legislators will always seek to “buy” support where it is cheapest, which directs them toward projects with the highest benefit-to-cost ratios. It makes no sense to offer $2 billion in tax cuts when the same goodwill could be purchased with a $100 million dam. An enterprising alliance might even pair funding with a $1.05 billion tax on the area’s residents—effectively splitting the gains between the locals and the central state. This revenue could then finance other priorities, whether to spur investment (with a corporate tax cut) or to reduce poverty (with an increase to SSDI). At no point does the welfare of the whole need to be considered: naked self-interest proves sufficient.

    • Absent some complicating pathology, a representative legislature will not support Dam 2. Spending $100 million to generate only $2,000 intensity of support from 25,000 people is inefficient. There are numerous other fractions of the population outside the coalition that could be courted at far lower cost through tax cuts or direct transfers.¹ This creates an implicit “ceiling price” for legislative support: any project that cannot generate at least break-even benefits will be displaced by a cheaper, higher-ROI alternative. So long as positive-return opportunities exist, coalitions will continue substituting toward them—capturing gains if necessary—until no support can be obtained at lower cost elsewhere. Projects exceeding this threshold, such as Dam 1 ($100 million cost for $2 billion in benefit split among 1 million), are adopted; those below it are not.

      • There might, however, be other countervailing forces working against efficient outcomes. A legislature may be representative along multiple dimensions (geography, demographics, party) but elected through single-member districts in which it is legislatively efficient for representatives to “soak” local minorities. (In America, that would be a policy disproportionately falling on blue-state Republicans and red-state Democrats to the advantage of red-state Republicans and blue-state Democrats.) Similarly, it is trivial to demonstrate that a legislator pointing a gun at their colleagues can extract additional concessions beyond what the above model will predict. Neither of these undermines the central argument; rather, they suggest what an exhaustive definition “representative” would need to include.

      • This is not an argument that, under representative conditions, the best possible outcome will always arise. Rather, it is a claim that there will be pressure towards a “very efficient” outcome. If one legislator is better at parliamentary procedure, then they could plausibly pull the package a small way away from “optimal” to the benefit of their constituents. The further from equilibrium the outcome gets, however, the more opportunities arise for mutually beneficial alliance amongst erstwhile rivals.

    • By contrast, when a legislature diverges from the population, it misjudges the benefits and costs of policies for that population. If the 0.25% of the population benefiting from Dam 2 is allocated 10% of legislative seats, the extent of its benefits is treated as equivalent to Dam 1. $100 million is cheap for 10% of the legislature, regardless of actual population representation. The further the composition deviates from the population, the more miscalibrated the legislative equilibrium becomes. Under these conditions, representatives maximize coalitional “return” not by finding the highest-value projects but by targeting the most overrepresented ones.

      • A central argument of Marc Reisner’s Cadillac Desert (1986) is that the federal government kept building dams in the Mountain West long after they stopped being enriching to society. If he is correct, the over-valuing of the local benefits relative to federal costs could be explained by the eight Mountain West states having 16% of the Senate but less than 4% of the population (c. 1960). This means that benefits to those states were “weighted” 4.5x more² than those to other states, making even inefficient projects there the cheapest way to “buy” votes.³

There are many ways to defend minority rights in a democracy—through courts, constitutions, or norms—but logrolling is the most powerful. Unlike other mechanisms, it is not at the whim of the majority’s caprice; instead, it embeds minority interests within the ordinary processes of governance. The ability of minorities to protect themselves from arbitrary treatment precedes and engenders the development of the other protective institutions. Seeing, say, an impartial judiciary as a prerequisite to minority rights gets cause and effect backwards. Only when the baseline equilibrium pulls in favor of equality do we have any reason to think judges would check the popular branches in favor of minorities rather than against them.

• Hypothetical 3: Imagine a society with two major economic sectors: 60% manufacturing and 40% agriculture. If legislative seats are allocated proportionally, one might fear that the manufacturing majority would simply soak the minority—say, by passing an inefficient tariff that raises manufacturers’ incomes by $60 million while reducing agricultural purchasing power by $100 million. The nation as a whole would be $40 million poorer, yet, since the costs fall entirely on the minority, the tariff appears to easily pass. This, however, overstates the vulnerability of a legislative minority. In practice, the agriculturalists need to only win over a small slice of the manufacturers to block the scheme. No coalition comprising a majority of society will be homogeneous, and so, while it cannot be known a priori which divisions inside the majority will prove decisive (e.g., geography, class, ideology, etc.), it is predictable that there is some cleavage which can be made politically salient.

  • Suppose there is a $10 million harbor project that would yield $20 million in benefits to the coastal manufacturers (roughly one-fifth of the total sector). This would mean that both “coastals” and agriculturalists would rather have the harbor pass than the tariff: such a scenario nets the former $6.8 million in additional gains, and the latter $96 million in avoided losses. This comes at the expense of the other manufacturers, who initially see a looming $64.8 million loss.⁴

    • The obvious response is for the manufacturers to simply fund the harbor themselves, and thus keep their coalition intact. But doing comes at the cost of shifting gains within the majority: when the coastals’ take rises (from $12 million to $30.8 million),⁵ the rest of the manufacturers’ share falls (from $48 million to $43.2 million). This both demonstrates the benefits of threatening defection at the same time as decreasing the benefits of loyalty. An ideological wing may now begin demanding reforms at odds with the rest of the party, or an ethnic constituency may insist on symbolic recognition. Such a dynamic—where concessions trigger further demands elsewhere—is why maintaining a majority becomes increasingly costly when it is oversized.⁶

    • Once the majority exhausts its positive-ROI projects, the minority remains able to propose redistributive bargains—some of which cannot be easily matched by the majority without fracturing it. The reason agriculturalists enjoy wider room for maneuver is that their preferred policy (canceling the tariff) is positive-sum, freeing $100 million for them while only costing their opponents $60 million. The agricultural minority might, for instance, offer a narrower $33 million transfer (without any positive ROI) to the coastal manufacturers, borne equally by all other regions. By defecting, the coastals nearly triple their gains (from +$12 million to +$33 million), the agriculturalists substantially cut their losses (from -$100 million to -$15 million), and the nation as a whole grows +$40 million richer as the inefficient tariff disappears. Of course, the non-coastal manufacturers are put down from +$48 million to -$18 million, a swing of -$66 million.

  • This dynamic—where minorities exploit divisions within majorities—naturally constrains arbitrary power. A 60% manufacturing chamber will legislate very differently from a 100% manufacturing chamber, even though “manufacturers can act unilaterally” technically describes both. The former must moderate to avoid exploitation of its internal divisions; the latter faces no such constraint. Proposed solutions that overrepresent vulnerable minorities inevitably backfire by underrepresenting other groups. In the U.S. Senate, rural Americans’ overrepresentation (the median state is 37% more rural than the nation) necessitates underrepresentation of racial minorities (the median state is 21% Hispanic or non-white versus 31% nationally). Overrepresenting agriculturalists in our hypothetical would simply create opportunities to exploit some manufacturing faction instead.

    • In real life, single-member districts do not produce representatives who are purely in-line with their district’s majority because a similar (but much less efficient) process of coalition-formation occurs pre-election. An oversized manufacturing majority will inevitably have internal disagreements, and the losing faction of the majority will have strong incentive to make a pre-election coalition with the minority. If our defecting manufacturer can moderate enough on industrial policy to make it clear that they are (to agriculturalists) the “lesser of two evils” they can still win as long as they retain at least 1/6th of the total manufacturers—likely by taking a polarizing stance on some other issue. Thus, some districts with 60% manufacturer voters are likely to still produce some relative moderates due to coalitions with agriculturalists. This does not, however, come close to replicating a proportional representation of agriculturalist interests in the legislature.

• If this seems like a weak safeguard for minorities, consider that every American belongs to some minority: Protestants make up 33% of the population, those aged 55–64 comprise 13%, Missouri residents 2%, secondary teachers 0.35%, and people with the last name “Wu” just 0.02%. Yet legislatures do not pass laws expropriating Protestants, the late-middle-aged, Missourians, secondary teachers, or people named Wu despite transparently being in the raw interest of “the majority.” This restraint flows from the logic of mutual vulnerability; being discriminated against is so unpleasant that most, if given the choice, would choose to refrain from discriminating against others if they were also protected. Exceptions arise only when a group is so stigmatized that association with them carries reputational cost. In the vast majority of cases, equal representation under universal suffrage is enough to secure minorities’ basic rights.

  • Late 19th-century North Carolina is a textbook case of minority logrolling. With a two-to-one white-to-black population ratio, the white supremacist Democratic Party consistently defeated the pro-black Republicans. Yet white North Carolinians were not a homogeneous bloc. Most notably, they were economically divided, with one quarter—overwhelmingly poor—defecting to the new Populist Party. The 1892 gubernatorial election indicates the balance of power: Democrats won just under half the vote, Republicans one-third, and Populists one-sixth. Republicans and Populists then “fused” before the next election, winning the governorship and legislature. The poor whites who followed their Populist leaders were not driven to this alliance by enlightened racial attitudes but by democratic arithmetic: the cheapest way to buy their liberation was to also concede it to their black allies. Thus, when “two wolves and a lamb” voted on lunch, having a vote prevented the lamb from being eaten.

    • White elites reacted by abandoning democracy. They first violently overthrew the biracial government and then disenfranchised the vast majority of black voters. Only once they were assured of a curated electorate did they resume elections—a state of affairs that would persist until the 1960s. Though horrific, this response indicates the power of minority logrolling: even one generation after the Civil War, Jim Crow could not be maintained democratically.

Our Legislature Is Not Representative

The idea that a legislature’s job is fundamentally one of representing the population goes back to at least the American founding. Here is John Adams, writing in the august year of 1776:

The principal difficulty lies, and the greatest care should be employed in constituting this Representative Assembly. It should be in miniature, an exact portrait of the people at large. It should think, feel, reason, and act like them. That it may be the interest of this Assembly to do strict justice at all times, it should be an equal representation, or in other words equal interest among the people should have equal interest in it.⁷

Since a legislature’s job is to create policy in the interest of all citizens—mindful of asymmetrical burdens but without over-weighting minority interests—it is crucial that it “should think, feel, reason, and act like” the people on whose behalf it is legislating. The composition of the U.S. House of Representatives offers an instructive test for how well-represented Americans are:

• 42% of America is non-white. 26% of House reps are.⁸

• 43% of voters are men.⁹ 71% of House reps are.¹⁰

• 38% Americans have a bachelor’s degree.¹¹ 94% of House reps do.¹²

• The median American is 39 years old.¹³ The median House rep is 58.¹⁴

• The median representative owns about as much as seven median Americans put together.¹⁵

One can make up stories to justify these discrepancies. Maybe we want our best and brightest to represent us, and 94% of our best and brightest completed college. Maybe we find older people with a longer track record to be more credible. Maybe we think wealthy people are less likely to be tempted by bribes. The racial and gender discrepancies cast doubt on such benign explanations (Do women actually think they are better represented by men?), but maybe—some might be tempted to argue—voters are okay downplaying demographic similarity relative to other factors. Who am I to tell a voter they should prefer someone who looks like them over someone who thinks like them? (It does, however, seem improbable that the former is usually a highly educated older wealthy white man).

But if voters were choosing to forego demographic similarity in order to maximize their ideological representation, then why would almost half of Americans have zero ideological representation? The House of Representatives is 50.8% Republican, 49.2% Democratic, and 0% independent of party,¹⁶ while 28% of voters identify as Democrats, 28% as Republicans, and 43% as independents.¹⁷ The House of Representatives does not well represent America either ideologically or demographically—and the Senate is even worse!

One must conclude that some (set of) mechanism(s) beyond voter preferences systematically shape who gets elected, producing a mismatch between the population and its representation. If democracy were merely a numbers game, one would expect at least a majority to find itself on the winning side; yet Americans are almost three times as likely to say that “their side has been losing more often than winning on the issues that matter to them.”¹⁸ This is a predictable result if, as this essay contends, the effectiveness of a legislature is contingent on its representativeness.

1. Note: It can be hard to target direct cash to one district physical place—even a cynical politician might hesitate before amending a bill to say “Residents of ID-2 get $2000 each.” Ben Nelson (D-NE) infamously tried to have the ACA pay for Nebraska’s (and only Nebraska’s) Medicaid expansion in the “Cornhusker Kickback” but that proved too brazen. (Alaska, though, does seem to get close to this sometimes.) Because of this, it is not necessarily the case that low ROI infrastructure etc. are never the best way to benefit a specific area, only that it is never the best way to win a comparably sized segment of the opposition—which need not be geographically concentrated.

2. (0.16/0.04)/(0.84/0.96) = 4.57

3. Robert Caro’s Master of the Senate (2002) extensively documents how Senate Majority Leader Lyndon B. Johnson found dams an especially useful way to extort votes on other issues (e.g., Civil Rights).

4. Coastals go from +$12 million (one-fifth of $60 million tariff windfall) to +$18.8 million ($20 million harbor windfall minus 12% of harbor cost), agriculturalists go from -$100 million (whole tariff loss) to -$4 million (40% of harbor cost), and manufacturers go from +$60 million to -$4.8 million (48% of harbor cost).

5. $30.8 million = $12 million (one-fifth of $60 million tariff windfall) + $20 million (100% of harbor benefits) -$1.2 million (12% of harbor cost)

6. William Riker’s The Theory of Political Coalitions (1962) is the canonical political science starting point for this concept. Chapter 3 (“Evidence about the Size Principle”) narrates how parties shed surplus supporters because maintaining oversized coalitions is costly—the titular “size principle.”

7. Thoughts on Government (1776): https://founders.archives.gov/documents/Adams/06-04-02-0026-0004

8. https://www.pewresearch.org/short-reads/2025/01/21/119th-congress-brings-new-growth-in-racial-ethnic-diversity-to-capitol-hill/

9. https://targetearly.targetsmart.com/g2024

10. https://cawp.rutgers.edu/facts/levels-office/congress/women-serving-119th-congress-2025-2027

11. https://www.census.gov/newsroom/press-releases/2022/educational-attainment.html

12. https://www.congress.gov/crs-product/R47470

13. https://www.nytimes.com/2023/06/22/us/census-median-age.html

14. https://www.congress.gov/crs-product/R47470

15. It’s hard to find Congressional net worth figures for after 2020, so I chose the most comparable citizen wealth number I could find (2019). For Americans: https://www.nerdwallet.com/article/finance/average-net-worth-by-age . For lawmakers: https://www.opensecrets.org/news/2020/04/majority-of-lawmakers-millionaires/ . I cannot imagine that the general ratio has shifted away from Congress members, as the S&P increased nearly 100% in the 60 months before October 2025, disproportionately enriching already well-off Americans.

16. https://www.pewresearch.org/short-reads/2025/01/21/119th-congress-brings-new-growth-in-racial-ethnic-diversity-to-capitol-hill/ https://en.wikipedia.org/wiki/List_of_current_United_States_representatives (accessed: April 15th, 2025)

17. https://news.gallup.com/poll/655157/gop-holds-edge-party-affiliation-third-straight-year.aspx

18. https://www.pewresearch.org/short-reads/2024/02/12/most-americans-continue-to-say-their-side-in-politics-is-losing-more-often-than-it-is-winning/