Major electoral reform is in vogue right now. Particularly from the left, calls for massive changes to the Senate, perhaps even its abolition, abound. The Electoral College is roundly loathed. House elections are condemned as hopelessly gerrymandered. We are made to believe that any mismatch between the popular vote and the results of an election is a threat to the very legitimacy of the American system and, implicitly, a kind of embarrassment to the country.
What is lacking, however, is any actual comparison to how things are done in other countries — and any analysis of what has changed since the days when the Left wasn’t complaining. And when we look at the evidence on how the American constitutional system is performing, we find that the U.S. looks pretty good compared with other countries or its own recent past.
House Elections Are Just as Democratic as Similar Races in Other Countries
We are sometimes given at least a few benchmarks, however. Some left-leaning critics compare the American system today to apartheid elections in South Africa. In those elections, massive gerrymandering, voter suppression, and disenfranchisement allowed a narrow clique of voters to control elections and oppress a dissenting majority.
So, is it the case that American elections are as lopsided as apartheid-era elections in South Africa?
The graph below provides a comparison. For U.S. House elections and South African legislative elections, it shows the gap between the winning party’s share of legislative seats won and that party’s share of the national popular vote. I’ll refer to this as the “representation rate” in the figures for simplicity. (This is a metric often used by critics of the U.S. system, though elections experts note that as a party’s popular-vote share rises above 50 percent, its seat share typically rises faster even in the absence of gerrymandering.) I use House elections because they are frequent and national, but will discuss Senate elections later.
As can be clearly seen, apartheid South Africa had a power mismatch two to five times the size of the one we see in the United States. The American experience simply has no meaningful resemblance to apartheid South Africa’s electoral imbalance.
But of course, the mismatch of votes and seats isn’t the only complaint against American elections today, or of South African elections during apartheid. In South Africa, non-whites were systematically locked out of participation. Many progressives claim the same about the United States, and particularly that GOP states try to implicitly or explicitly disenfranchise many voters.
We can do a crude test of this theory by switching from a party’s share of votes cast as our benchmark, to a party’s votes as a share of national population. I’ll refer to this gap as a country’s “democratic deficit” in the figures. This is skewed of course by age composition and the presence of noncitizens, but it’s a useful enough first pass. If the gap between the share of seats won and the share of the population that voted for a party is larger, that means a narrower slice of the population is ultimately in charge of the country. Here, the difference between the U.S. and apartheid South Africa becomes even clearer.
The difference between share of seats won and the share of the population to have voted for the winning party was enormous in apartheid South Africa. Voters amounting to about 3 percent of the population were able to capture about 60 percent of the legislature. That is what a dysfunctional, non-democratic system looks like: a 57-point gap. Critics of the American system should keep a little bit of perspective in mind.
In the U.S., at the absolute worst in recent history, voters representing 14 percent of the population captured 60 percent of the House seats. That was in 1974, when Democrats took 291 seats in the House. Indeed, whichever metric you use, the last 20 years have seen a considerably more “accurate” outcome for the House than the 20 or 30 years before that when Democrats held durable control. It is convenient, then, that Democrats have finally woken up to the mismatch between votes cast and seats won.
Of course, “not as bad as apartheid South Africa” is a really low bar. The point progressives are really trying to make is that nobody in respectable countries, such as Canada, or Sweden, or Germany, would ever tolerate our terribly undemocratic system.
But it turns out that comparing ourselves with other developed countries still doesn’t make the United States look very bad. Here’s the same metric as the first graph (share of lower house seats won minus share of popular vote for lower house elections), for a wider range of countries:
Simply in terms of lopsided favoritism for winners, many other developed countries actually are as bad as apartheid South Africa was. The United Kingdom, France, Japan, and Italy have all seen elections give the kind of lopsided results that were observed under apartheid. The U.S. isn’t the best of all democracies, but we are nowhere near the worst. And if current results from the 2018 elections hold up, they will have been one of the closest approximations of the national popular vote we have ever had (2006 and 1994 also saw very close approximations).
In fact, the United States has substantially less lopsided lower-house elections than most countries have. Whatever our gerrymandering problems may be, the political system of, for example, France, is far more dysfunctional. Macron’s party won nearly 60 percent of the legislative seats with just 40 percent of the vote. Trudeau’s Liberals in Canada aren’t much better: They won about 55 percent of seats with about 40 percent of the vote. Of course, a country with proportional representation, such as Sweden, does better, but even then there’s still a gap due to threshold rules and other idiosyncracies of the election system. Indeed, the U.S. election system tends to produce outcomes very close to those observed under proportional representation in Sweden, especially in the 2018 election.
When we turn to look at share of seats minus share of total population won, the story gets even more interesting.
Here, it becomes clear that the U.S. does perform a bit less democratically than peer countries do. A smaller share of the total population ultimately votes for the winners. We can also see what made apartheid elections so unusual: not the gerrymandering, but the extraordinarily low electoral participation. Because turnout is lower in the U.S. than in many foreign countries, our political victors tend to have received votes from a smaller share of the population.
But again, the U.S.’s position here is at worst a mediocrity, not a catastrophe. Our winners tend to have a relatively smaller base for their mandate than do politicians in Germany or Sweden, and recently the U.K. But still, the gap between share of population won and share of seats won is larger in France and Japan than in the United States, and our outcomes are very comparable to those of our near neighbor Canada.
Which is all to say, as far as lower-house elections are concerned, the U.S. is at worst about average in terms of how closely our results fit the popular vote. To the extent we are an underperformer, it’s really just about low turnout.
But while progressives often criticize the House for its gerrymandering, their main ire is usually reserved for the U.S. upper house: the Senate.
The Senate Represents the Population as Well as Most Developed-Country Legislatures Do
We can calculate the same figure for the Senate as for the House; it’s just a bit more complicated. Because only a third of senators are elected each year, we have to take the rolling vote total of the last three elections for senators as the implied “senatorial popular vote,” and the rolling total of seats won of the last three elections for senators as the implied “senatorial seats won.” This gives us an even balance of states. Of course, there’s no actual national popular vote for the House or the Senate, and local variation in what’s on the ballot, as well as local candidate-specific factors, all influence how races go; but to counter progressive arguments, we have to use the data we have at hand.
You might expect that the Senate would have a huge gap between share of seats won by the winner and their share of the popular vote. But it turns out that the Senate has historically been quite balanced. The graph below shows the vote share minus seats won share for the U.S. House and Senate, and the average of developed countries discussed above.
As you can see, the Senate has historically been a more direct approximation of direct democracy than is the typical rich-country legislature in my sample. Now, in 2018 things have gotten a bit lopsided in the Senate, with Republicans controlling an outsized share of seats. However, with 2020 shaping up to be a rough year for Republicans given the election map, the mismatch will probably tick back down.
Far from being hopelessly lopsided against the popular vote, the Senate has historically been nearly as reflective of the popular vote as the House. The last few years have seen a bit of a deviation, but the long-run trend seems to be similarity between the two.
We can compare as well how turnout affects these metrics, as we did above.
Again, the House and Senate are basically equivalent here. The House oscillates more, but the trend is virtually identical. Owing to low turnout, winners of U.S. Congress seats tend to be elected by a smaller share of the population than the winners of elections in foreign countries, but the Senate is no more problematic in this regard than the House. For both the House and the Senate, it takes about 15 to 20 percent of the U.S. population voting for a party to give that party a majority.
On the whole, U.S. legislative elections are well within the normal experience of developed countries in terms of the share of seats won compared with the share of votes won. The only place where the U.S. system truly appears a bit exceptional is in our low level of turnout.
American Elections Aren’t Very Biased
But many progressives won’t find these arguments convincing. “That’s all very well and good,” they might reply, “but our argument isn’t that the U.S. system is fundamentally anti-majoritarian in the aggregate, just that it reliably tilts in favor of one party, whether that party has a majority or not.” This argument is that the U.S. system isn’t so much anti-democratic as it is anti-Democratic; more Republican than it is republican.
Conveniently, this claim is also something we can check in the data. We can look over the past several election cycles and see how many seats Democrats and Republicans received in the House and the Senate compared with how many votes they got. Basically, by looking at the Democratic party’s popular-vote share in House and rolling Senate elections versus share of seats won, and doing the same for the GOP, we can see if our system systematically favors one or another party.
The chart above shows pretty clearly that there is currently a Republican bias in our electoral system: The GOP gets more House and Senate seats than its popular vote share would imply (although notably, in 2018 the gap in the House essentially disappeared). Point, progressives.
But hold on. That’s a quite new phenomenon, and indeed the Republican House bias at its peak was still smaller than the bias in favor of Democrats was in the early 1990s. The Republican Senate bias is quite large, but again, is not radically outside the U.S. historic norm.
In other words: The U.S. partisan bias today is not particularly severe in historical context. What is different about it is that, after decades and decades of a built-in structural advantage for Democrats, the tables have begun to turn. The response from the left has been to announce that there is a constitutional crisis and that the whole system must be scrapped.
In reality, the American political system is no more structurally biased today than it has been in the past. It’s just a different group of people benefiting. In the future, the tables will turn again. There will come another time when the Democratic coalition will be aligned more favorably with respect to American political geography. These transient biases are no reason to make permanent changes.
But the key question remains: Are these biases large or small as compared with those in other countries? Tracking the whole partisan mix of other nations is very labor-intensive, so I can include only a more limited set of countries. But I have classified parties in the U.S., Canada, France, Sweden, and Germany as left, right, or neither. For example, in France, parties of the parliamentary right, as well as nationalist parties, I classify as of the right. Socialist, Communist, far-left, green, and parliamentary-left parties are all classified as the left. Meanwhile, Emmanuel Macron’s En Marche party is neither. My interest is in the net bias in favor of right-wing parties over left-wing parties: basically, a measure of how much the electoral system over- or under-awards legislative seats to rightists versus leftists.
As you can see, the U.S. has a very normal amount of bias. Around 2008 to 2010, we actually had the most left-tilted electoral system in this admittedly small sample. In 2016 and 2017, of course, the U.S. had a very right-tilted system. But that’s a pretty recent thing, and as recently as 2014 to 2015, our system was approximately as right-biased as Sweden’s, while in 2018, the bias shifted leftwards a bit. That’s not exactly a stinging indictment. Our system is not appreciably more biased than those of other developed liberal democracies.
Beyond the numbers, the question of a structural advantage for one party or another is a bit odd. What kind of “structural” advantage or inherent bias is so fickle that in 2004 the net bias is eight percentage points in favor of Republicans, then in 2008 is six percentage points in favor of Democrats, then in 2016 is twelve points in favor of Republicans?
It All Comes Down to Turnout
The point is, these advantages aren’t really structural. They’re about the very specific, time-varying idiosyncracies of electoral coalitions. Persuade a few more people in swing districts, and your party can pick up a lot of seats without adding much to your vote share, and thus obtain what looks like a “structural” advantage. Whenever you hear that Republicans or Democrats have some inherent advantage or disadvantage, just remember that the long and variable history of U.S. elections shows that this isn’t true: Electoral biases bounce around quite a bit, year to year, even.
In other words, “structural advantages” all come down to turnout. And speaking of turnout, I’ve already mentioned that, while the U.S. is quite normal in terms of the extent of gerrymandering or structural advantages and disadvantages, where the U.S. is exceptional is in our low election turnout.
Some sources compare election turnout internationally using the voting-age population by country. Others use registered voters. Both methods have flaws: Using the voting-age population misses that countries vary widely in how many adults are citizens; countries that have more non-citizen immigrants, like the United States, will look like they have lower turnout than countries with fewer non-citizens, like Japan. On the other hand, using registered voters as the denominator also has flaws, as countries vary widely in their method of registering people to vote. The result is that a country like the United States, where voter registration rates are low, will look like it has better turnout than a country where voter registration rates are very high, like the Netherlands. Using the number of adult citizens in a country is the best benchmark for assessing the scale of electoral participation.
Of course, it’s also a very hard metric to use, as most countries make it rather difficult to track down data on the number of people in a given year who were over age 17 and citizens. I have managed to assemble a comparison of 27 countries, but even that list requires a substantial amount of extrapolation and imputation. Still, for all the faults in the data, this method is probably still better than comparisons based on registration rosters or the voting age population.
As you can see, the U.S. has one of the lowest rates of voter participation of any developed country in my sample. Only Japan, famous for its uncompetitive elections, is lower. There are several developing countries with lower electoral participation in recent elections, but on the other hand, many quite poor countries still manage to beat the U.S.
Meanwhile, Belgium, Australia, and New Zealand are standouts for high electoral participation. Australia and Belgium, of course, have compulsory-voting laws, as do Turkey and Brazil, two developing countries with high turnout. New Zealand is a bit of a mystery to me; it has turnout comparable to compulsory-voting countries, yet no compulsory-voting law.
This, then, is a place where the U.S. is truly exceptional: Our citizens are simply less likely to behave like citizens and participate in governance than those of other countries. There is debate about why this may be, with proposed explanations including the workday timing of U.S. elections, apathy arising from the two-party system, active voter suppression efforts in some states, the prevalence of felony disenfranchisement, and the excessive frequency of U.S. elections and the many things we expect voters to vote on. But it may be worth noting that turnout rates vary appreciably within the United States as well. The map below colors each state by its average turnout rate among adult citizens in the presidential elections of 2000, 2004, 2008, 2012, and 2016. I use the average to avoid the skewing effects of occasional Senate or gubernatorial races.
States like Minnesota and Maine have turnout rates at parity with the rest of the developed world: 70 percent or more of citizens vote in a given election. Meanwhile, states like West Virginia, Arkansas, and Texas have much lower turnout, around 50 percent. Even states like California and New York don’t do that well, coming in well under 60 percent. Turnout differences don’t have a tidy partisan bias we can point to. Even states with extensive early voting, numerous polling places, proactive registration campaigns, and competitive elections have lower turnout than our rich-world peers.
But if activists want to make the American political system more representative, they should spend less time worrying about the rural bias of the Senate and gerrymandering in the House, and more time asking why so few Americans show up to vote. And on that, I would note another place where American governance is truly exceptional.
The United States has enormous legislative districts, with the result that citizens are extremely disconnected from legislators. The only country I could find with bigger districts, not shown here, is India. Increasing the number of representatives by a large amount would almost certainly increase political participation by creating smaller districts that could more plausibly be won by local elites without national backing. More to the point, small areas tend to be more politically idiosyncratic than large areas simply due to the law of averages, and thus smaller districts would probably yield politicians with more diverse policy views. Other countries with single-member district elections like we have, such as Canada, nonetheless manage to have much greater political diversity represented in major parties, perhaps thanks to having much smaller legislative districts.
Expanding the size of the House of Representatives is a good idea for many reasons, not least that it would probably help increase political participation. But it wouldn’t fix all our problems. Figuring out what else, if anything, should be done about low turnout is a question for another day.
For now, it’s enough to stop here. As I have shown, American politics today are no more anti-democratic or anti-republican than they have been in the past. Our elections yield results pretty closely aligned to the popular vote, and indeed our elections actually yield more “accurate” results than the elections of many countries progressives would have the U.S. emulate.
The Left isn’t furious now because it actually has some principled belief that the share of seats won should always match the share of votes won, but because, after decades of the American political map favoring the Left, the map has finally begun to favor the Right a bit. In time, the tables will turn again, and the Right will face a “structural bias” — and in all likelihood, some conservatives will complain about the unfairness of the system.
Alas, people are imperfect. Luckily, our system depends not just on people, but on a written, binding, permanent set of rules, our Constitution. Under these rules, sometimes one side has a leg up, sometimes the other side does, but the system, through all these storms, is strong.