U.S. democracy is ranked 25th, let’s fix that
Introduction
The late congressman John Lewis wrote that “democracy is not a state. It is an act.” Two and a half centuries ago, we built our country on the idea of democracy, and while we have followed Lewis’ ideal by adapting and improving ever since, there is still more we can do. When ranked according to five key measures of democracy by the Economist — “electoral process and pluralism, the functioning of government, political participation, democratic political culture, and civil liberties” — the U.S. places only 25th, down from 17th ten years ago.
Implementing the following policies would help to make U.S. elections and governmental representation more fair and more democratic:
- Pass the National Popular Vote bill in outstanding states to have the electoral college vote for the winner of the national popular vote.
- Give Puerto Rico and D.C. the opportunity to become states (or independent for Puerto Rico).
- Consider where there might be better methods than first-past-the-post voting to use in elections.
Improving the electoral college
100 senators + 435 representatives + 3 votes for D.C. = 538 electors who choose the president of the United States. However, depending on which state you live in, your vote may be worth more or less than someone’s from a different state — not everyone’s vote for president is equal.
What is the problem with the electoral college?
Problem 1: The electoral college under- and over- represents certain populations. A California elector represents ~720k people, but a Wyoming elector represents only ~190k. The founders chose to design our presidential elections so that small states would not be ignored, but states aren’t the only way to segment our country. There are many other cross-sections we can take of the U.S. besides by state, which don’t receive the same attention to small voices.
The figure¹ below shows a few different ways to segment America, by population density, race, Hispanic origin, income, education, and health insurance type, along with how those segments are over- or under-represented by the electoral college. The left column shows the current population breakdown. The next column shows the difference in how electoral college votes represent the U.S. population from the true population. The two right columns then illustrate the idea of the higher importance of swing states by leveraging a couple metrics — voter power index and tipping point probability, defined in the figure’s footnote — from FiveThirtyEight to show how the electoral college represents the part of the population most likely to matter in an election.
The electoral college amplifies rural, White, and Medicare recipient (elderly) voices, and mutes Black and Hispanic, high income earner, college (bachelor’s degree) graduate, and uninsured voices.
Problem 2: Not everyone’s vote matters. As seen in the right two columns in the figure above, voters in strongly partisan states are unlikely to impact the national results. For example, even through ~79% of voters in Effingham County, IL favored Donald Trump, they were far outweighed by the ~74% of Cook County (Chicago) residents who voted for Joe Biden. Similarly, the ~71% of voters in Travis County, TX (Austin) who favored Biden were outweighed by the red suburban and rural parts of Texas².
This also impacts presidential campaigns, and in some cases, even policy. Candidates spend more time and money on swing states, ignoring those who live in others. On the policy side, the distribution of federal funds can favor swing states³, and we saw the issue of fracking take front stage in the 2020 debates, not because it’s a significant issue nationally, but because it is significant to swing voters in western Pennsylvania.
Problem 3: We’ve had 5 presidents elected without winning the popular vote. This has now happened 2 out of the last 6 elections (2016 and 2000), and also occurred 3 times in the 1800s. Hamilton argued in Federalist No. 68 that the electoral college results, if different from the popular vote, would protect against a less “capable” public selecting an unqualified candidate. However, it seems nearly inconceivable in today’s America electors would take it upon themselves to change the results of an election.
The results of 2000 and 2016 were not due to electors acting with greater “information and discernment” to choose the president, but were due to how electoral college votes structurally differ from the popular vote. In other words, these results were not the intentional product of the electoral college design, but an inadvertent deviation from the majority based on how the U.S. population happens to be distributed today.
Problem 4: We don’t think in terms of states like we used to. The idea of ensuring all states an equal vote came from a time when strong state-by-state political differences existed. Rhode Island was politically distinct from Massachusetts, which was different than New York, and so on. But today, more so than in the 1700s, a Republican voter in Georgia is likely to hold similar views to a Republican voter in Oregon. Same for a Democrat in California and one in Maine. Our political views are less driven by the state we live in.
Some proof of this: The Economist developed a demographic-based voting predictor model through surveys in 2018. The model shows that a person’s geographic region is, on average, the 7th best predictor of partisan lean, behind religion, race, sexual orientation, level of education, population density, and gender. How does this compare to how important it was in the 1700s? Well we don’t have survey data, but you can cut out race and gender based on the electorate, sexual orientation based on the era’s politics, and you could make a strong argument that the impacts of education level and population density were likely less predictive of partisan lean based on the relative homogeneity of the electorate (white land-owning men). Maybe there are other demographics that were important back then that we could add, but this look gives us a quick insight into how state-by-state and regional differences in voting have very likely become tertiary in importance to other segments of America over time. If this is the case, then why are we using a system that goes out of the way to equalize state voices at the expense of one person, one vote?
How would a national popular vote solve these issues? And how do we enact it?
The most democratic way to solve these issues is to elect the president purely by popular vote. Groups would no longer be under- or over- represented as compared to their population, everyone’s vote would matter equally, the president would always win the popular vote, and we’d update our election system to reflect modern national-level (rather than state-by-state) political thinking.
Federally, a national popular vote could be enacted by an amendment, but would be difficult to pass. However, this can be solved at the state level via the National Popular Vote Interstate Compact, which is an agreement among states to allocate all of their electors to the national popular vote winner, rather than the winners of the individual states. The compact is passed as a bill in state governments, and will not take effect until enough states join to comprise a majority of electoral votes.
But wouldn’t this…
(Addressing a non-exhaustive list of common concerns)
Benefit the Democrats? If you look at numbers from past elections, then this may look to be the case. However, we’re unlikely to end up with Democrat after Democrat under a popular vote, especially not long-term, as there are a couple factors to keep in mind:
a) Presidential campaign strategy and policy is effectively dictated by splitting the country in half, which we see in the constant back and forth between parties for the presidency. If one party is underperforming the other, then that party will adjust their platform slightly to capture swing voters. In other words, the parties operate a lot like the free market, adjusting themselves to find the middle ground (the Nash equilibrium), so that they can hold their base while obtaining enough votes to win. Parties will adjust, and future elections will remain contentious. We can see this constant flux back and forth in the figure below, which shows the popular vote margin in House elections. House data was used to minimize the impact of individual candidates to assess party dynamics only. The popular vote margin for either party has not exceeded 10% since 1976 (start date based on data availability), and has, following a large margin one way, often swung strongly back the other way 2 years later. Party control of the House has also been roughly equal through the period.
b) State-level political lean is in constant flux. An election system that benefits the Republicans today may benefit the Democrats tomorrow. While we often think of the electoral map as set in stone, California voted for Bush in ‘92, and Texas voted for Carter in ‘76, neither election all that long ago in political time. In fact, it’s widely believed that it’s a matter of time until the electoral college favors democrats due to demographic shifts in Texas⁴. We’ve seen similar blue shifts in Virginia (Blue since 2008, had been red since LBJ in ‘64), North Carolina (Now a swing state, had been Red since ‘76), Colorado (Blue since 2004, had been Red since ’64 with the exception of ‘92), and Arizona (Blue in 2020, had been red since Truman in ’48 with the exception of ’96 ). Conversely, similar red shifts have been limited to Ohio and Iowa, threatening to turn the electoral college structural advantage over to the Democrats.
Lead to only the major cities deciding the president? Probably not, but I’ll give a nuanced answer. To understand if a popular vote would result in cities “deciding” the presidency, let’s look more closely at a) the voting population in major cities, b) how easily moderate urban voters can be swayed, and c) how many votes can be gained by increasing turnout in cities.
a) On population, America’s big cities are smaller than many may think. Looking just at city population, the top 10 cities only account for 8% of the U.S. population. The numbers below show that most of America resides in the suburbs of medium-sized cities, from Boston (#11) to Birmingham (#50)⁵. There aren’t immense numbers of voters from New York, Los Angeles, and Chicago who will dictate the whole election.
b) As for swaying moderate voters, let’s consider the swing from 2016 when congress was fully red to 2018 when the house flipped blue to see if that moderate swing was from cities or elsewhere. The Pew Research Center found that in comparing urban / suburban / rural votes from 2016 and 2018, the suburbs were responsible for the largest blue swing. The 9 point swing among suburban voters (53% of electorate by Pew’s definitions) was greater than 2 points among urban voters (23% of electorate) and 4 points among rural voters (24% of electorate). The suburbs have more people, and are more swingy. It’s possible this is a factor of turnout among partisan voters living in suburbs, but either way highlights a greater suburban than urban influence. Furthermore, the Economist model I mentioned earlier ranks population density as the 5th greatest determinant of someone’s vote. It’s more “who” is deciding the president among moderate voters than “where.”
c) Finally, on voter turnout, using 2016 numbers, cities do appear to have lower turnout (calculated as votes / total population, due to data availability) when regressed on county population density (log variable) alone, though the variable is not highly significant with a p value of 17.8%. The results of this (logit) regression show that a county with 1000x the average county population density has a predicted ~4% lower turnout than a county with 1/1000x the average population density⁶.
This result implies there may be a slightly larger “get out the vote” opportunity in urban areas, which lean left, but only if “get out the vote” efforts are effective in those areas. However, back to the data from (b), keep in mind that this “get out the vote opportunity” is hypothetical from the data above; the actual results we’ve seen are swings from the suburbs.
As I concluded above, urban areas probably would not end up deciding the election entirely, but the answer is undoubtedly murky. This could also be just a short term dynamic. For all we know, the 2040 election may not be urban vs rural. Maybe it’s West vs East, or oil states vs clean energy states, rendering this concern moot entirely.
Cause candidates to only focus on / campaign in the major cities? This is a reasonable concern, but similar to the above answer, most likely not. As evidence of this, we can look within the states themselves, where a popular vote decides the winner of the state’s electors. Do candidates only focus on the cities there? Borrowing more data from the national popular vote website, of 618 general campaign events in 2012 and 2016 that took place in the 12 major battleground states, 31% of those events occurred in the largest metro areas, where 30% of the population resides. So there was a delta of just 1% between population and campaign event frequency in urban areas. Maybe campaigns look very different in this new reality, but the data tells us it’s unlikely that candidates will only focus on major cities.
Cause candidates to ignore small rural states? Quick answer to this one. Under the current swing state paradigm, rural states like Wyoming, Kansas, etc. are strong red and don’t have a major impact on the general election outcome. These states are more or less ignored today, and a popular vote would not change that.
Statehood for Puerto Rico and D.C.
Before discussing Puerto Rico, it needs to be noted that there is significant debate within Puerto Rico around the island’s future. As the ultimate decision should be up to the Puerto Rican people, I argue here that the U.S. should provide the opportunity for Puerto Rico to vote on statehood, independent of the benefit or loss of any political party or group. The people should also have the option for independence, or other statuses, available to them as well in this vote.
Puerto Rico has been a part of the U.S. since the end of the Spanish-American war in 1898. The island’s government today is a commonwealth, operating similar to a state, but with a few key differences. Most notably, these include lower Social Security and Medicaid benefits, barriers to the island filing for bankruptcy, the exemption from federal income tax, and no representation in congress or the electoral college. The island is home to ~3.2M people, which is greater than the population of 20 states.
The political status of the island has been subject to frequent debate. Options discussed include remaining a commonwealth, statehood, independence, and free association. In the 2020 election, 52% of voters responded “yes” to the non-binding question “Should Puerto Rico be admitted immediately into the Union as a State?” Additionally, both the U.S. Republican and Democratic parties’ official platforms support Puerto Rican statehood⁷.
So, what’s the problem with Puerto Rico’s current status?
Without federal representation, Puerto Ricans are unable to have a voice in the federal laws that affect them. These include, but are not limited to, Supreme Court nominations and decisions, non-income federal taxes, Social Security, and healthcare.
But wouldn’t this…
Also benefit the Democrats? Maybe — let’s consider a couple different factors when looking at this:
- Many politicians assume Puerto Rico would vote Democrat; however, island’s perceived blue lean is actually difficult to assess. Unfortunately, there aren’t major polls to give us a definitive answer. Although there was 2x turnout in the 2016 Democratic presidential primary versus the Republican primary, only 4% of the population voted, so this isn’t the best indicator. The island does have a non-voting representative in the House, but Puerto Rico’s political parties are separate from the Democrats and Republicans. The commonwealth’s current representative, Jenniffer Gonzalez, actually aligns herself with the Republican Party — so while a slight blue lean may be theorized, the island’s representative today is red. It’s by no means a guarantee that Puerto Rico would hand the Democrats two extra senate seats.
- Future politics could look very different, as mentioned in the electoral college section above. So even if Puerto Rico ends up leaning blue today, that could very well be just a short-term dynamic.
Hurt the federal government by assuming Puerto Rico’s debts? Exacerbated by the impacts of Hurricane Maria, the Puerto Rican government faces a ~$70B debt crisis⁸. As statehood would have the federal government assume this debt, is this a significant problem?
While ~$70B seems like a large number, it is <0.25% of the $27T federal debt, less than the 2018 debt of 11 states (CA, NY, TX, IL, FL, PA, MA, NJ, WA, OH, MI), and comparable to that of individual large cities (NYC at $186B, Chicago at $34B). The relative magnitude of Puerto Rico’s debt and economy would likely render the impact on the federal government small to negligible⁸.
What about D.C.?
Basically, the same argument here. There are 700k people who live in D.C. (greater than the population of Vermont and Wyoming) without congressional representation, but subject to federal laws. Since D.C. is smaller than Puerto Rico, has electoral college votes, and has a far stronger and clearer partisan lean (voted 92% for Joe Biden in 2020), D.C. may be more politically contentious than Puerto Rico. However, if we seek policy that prioritizes democratic values over partisan values, the answer on D.C. is also the same: statehood.
A few additional thoughts on voting systems
I was originally going to use this final section to argue for an alternative to our first-past-the-post (FPTP) voting system, in which the candidate who gets the most votes wins, irrespective of the secondary preferences of voters. FairVote, an advocacy group, has recently garnered momentum advocating for instant-runoff voting (IRV, also called ranked choice voting), including Maine implementing the system for all state and federal elections in 2016. However, IRV wouldn’t necessarily be an improvement on FPTP.
The difference between IRV and FPTP can be demonstrated with an example: Let’s say there are 3 candidates, A, B, and C. Candidate A receives 40% of the vote, B 35%, and C 25%. Under FPTP, candidate A wins, they have the most votes. But what if all the voters for C prefer B over A? B would have been preferred to A head-to-head, but they lost votes to candidate C. In IRV, voters would rank their preferences, the last-place finishers would be eliminated, and their votes would be reallocated to the second place candidates on those ballots. This process repeats until one candidate has >50% of the vote, which would be candidate B in our example. Wikipedia is also helpful here in comparing the two systems, in particular through an example of voting for the location of the Tennessee state capital here and here.
As illustrated in the example above, FPTP does not always pick the “most desired” candidate. We see this from observing recent election results, and from the lack of third party success in the U.S. In the 2000 presidential election, many would argue that votes for Ralph Nader in Florida, whose supporters generally preferred Gore over Bush, prevented Gore from winning the presidency. Similar impacts of “spoiler candidates” were also seen in the 2020 and 2016 presidential primaries. Biden’s Super Tuesday surge was in large part fueled by the consolidation of the moderate wing of the Democratic party, with Buttigieg, Klobuchar, and O’Rourke all dropping out to back Biden, allowing him to win the FPTP election. The 2016 Republican primary was a different story, in which the moderate faction of Rubio, Kasich, and Bush cannibalized itself, allowing candidates from other factions of the party — Trump and Cruz — to pull ahead.
Would IRV be better than FPTP? Maybe. Electoral systems can be analyzed mathematically, assessing how different methods of voting satisfy a variety of criteria that we would deem important in an election. For example, could the presence of an “irrelevant” (spoiler) candidate change the result? Could the winning candidate defeat all other candidates individually in a head-to-head election? Unfortunately, a concept called Arrow’s impossibility theorem more or less states that no electoral system can satisfy all the criteria we would deem important. As a result, the chosen electoral system varies significantly around the world.
So what’s the answer? Well, it depends. There’s no perfect system, so we have to evaluate trade-offs among the different options, which are laid out rather clearly here. The “best” system would balance the criteria we deem most important to satisfy, the criteria we would deem most outrageous if violated, the types of “odd” outcomes we could see (e.g., spoiler candidates), and how likely those odd outcomes are. The answer would likely vary by type of election as well, what works best for the senate may not work best for mayor.
I am neither going to support nor oppose FairVote here, but since the IRV system appears to have the most momentum today, I would encourage you to consider the options and tradeoffs between IRV and our current system. Any other approach is most likely to be confined to the realm of academic papers, at least for now.
Currently, our federal government is not accurately representing the people as best it could, and all three of these policies work to remedy this issue. I’m not advocating for any novel or groundbreaking policy here, but I wanted to present these ideas in as non-partisan a fashion as I can, supported by a variety of facts and some custom analyses. Maybe I change your mind with new data and arguments to consider, or maybe I don’t, but I do hope you take some time to think through the information presented here and see if that brings you any new ideas, questions, or conclusions.
¹ U.S. Census; KFF; DQYDJ; FiveThirtyEight; Custom analysis
² New York Times election data
³ National Popular Vote; https://www.vanderbilt.edu/csdi/research/CSDI-WP-01-2011.pdf
⁴ San Antonio Express-News; Texas Tribune; Vox
⁵ U.S. Census
⁶ MIT Election lab; U.S. Census
⁷ 2020 Republican Party platform; 2020 Democratic Party platform
⁸ NBC; U.S. Treasury; Truth in Accounting; Statista; BEA
