The Myth of Democrats' 20-Million-Vote Majority
This past weekend, Vox’s featured piece contained the click-bait-y headline “The Senate's 46 Democrats got 20 million more votes than its 54 Republicans.” The short, three-paragraph piece concludes:
“This doesn't mean that the Republican majority is illegitimate or anything like that. Indeed, after 2008 and 2012, the tables were turned: Democrats got more Senate seats than their vote share suggested they should. The problem isn't that the deck is stacked in favor of Republicans. The problem is that the deck is stacked in favor of small states, which receive equal representation in the Senate despite dramatic variance in population. The Senate is a profoundly anti-democratic body and should be abolished.”
This is flawed, badly, for four reasons. (We’ll ignore the concluding sentence, if for no other reason than that Article 5 of the Constitution prohibits relitigating the Connecticut Compromise, even by amendment.)
1. The 20 million figure is misleading.
I’ll admit it took me a while to figure this out, but the “20 million votes” statistic isn’t what author Dylan Matthews implies. One might get the impression from Matthews’ article that Republicans lost the popular vote overall by 20 million votes, especially since Matthews refers to “vote share” in his concluding paragraph. Worse, Matthews claims that a chart comparing percentages of votes won by each party to seats won represents the 20 million discrepancy “in chart form.”
If Senate Republicans had lost the popular vote by 20 million votes, yet won 54 percent of the seats, it would be problematic indeed. To put this in perspective, Barack Obama won by just shy of 5 million votes in 2012. Democrats winning the popular vote for the Senate by 20 million votes -- 67.8 million to 47.1 million, according to the article -- would be a popular vote win equivalent to Ronald Reagan’s margin in 1984. One imagines Republicans would be a bit peeved if Reagan had won by that margin and yet had lost to Walter Mondale.
But the 20 million statistic isn’t at all about vote share. This is just a total of the number of votes that victorious Senate Democrats won, compared to the number of votes claimed by victorious Senate Republicans.
The fact that the 46 Democrats got more votes than their 54 Republican opponents isn’t “crazy,” to borrow Matthews’ term, at all. Nor does it tell us much of anything about malapportionment in the Senate. The reason that victorious Democrats won 20 million more votes than victorious Republicans is that half of the victorious Democrats won in 2012, while 80 percent of the victorious Republicans won in 2010 or 2014. Presidential elections have higher turnout than midterm elections. So a win in 2012 will necessarily bring more votes than a win in 2010 or 2014, even in a similarly sized state.
For example, Heidi Heitkamp barely won her Senate seat in 2012 with 161,000 votes, for a one-point victory. Two years earlier, her co-senator, John Hoeven, won with over 70 percent of the vote. But because 80,000 fewer votes were cast in 2010 than in 2012, Hoeven received only 182,000 votes.
If we look at other states with mixed delegations and mismatched elections, we see the same thing. Marco Rubio won a three-way race in 2010 with 2.6 million votes representing a win of almost 20 points; Democrat Bill Nelson won in 2012 with 3.5 million votes in a 13-point win. Republican Dan Coats of Indiana won by 15 points in 2010 while Democrat Joe Donnelly won by just six points in 2012, but Donnelly is credited with 1.3 million votes and Coats just 950,000. Democrat Claire McCaskill and Republican Roy Blunt both won by about 15 points in Missouri, but Blunt got just 1 million votes to McCaskill’s 1.5 million (incidentally, this is the same 2:3 ratio we see in the overall numbers). Democrat Jon Tester of Montana won by four points in 2012 with 236,000 votes; two years later Republican Steve Daines won by 18 points but received just 211,000 votes. Democrat Sherrod Brown of Ohio won by six points with 2.8 million votes in 2012; Republican Rob Portman won by 18 points in 2010 with just 2.1 million votes.
There are a handful of states where the relationship worked in reverse (Dean Heller of Nevada won more votes than Harry Reid despite the latter’s larger margin), but overall there are huge discrepancies between vote shares won and the number of votes won even among senators from the same state. They almost all work in Democrats’ favor since 2012 was such a good year for them, and the votes cast that year were more numerous than the votes cast in 2010 and 2014.
If you are interested in what vote share can tell us -- and we’ll focus on this for the rest of this article -- we can just add up the total number of votes cast for all Democratic candidates and Republican candidates in the 2010, 2012 and 2014 elections, using Dave Leip’s election atlas as a source. It works out to 103 million votes for Democrats to 98 million votes for Republicans: a difference of 5 million votes, or 2.5 points. This is still a meaningful difference between the share of votes cast for Democrats and the share of seats that they won, but it isn’t the outrageous result that Vox headlined.
Even this overstates things, though. If you eliminate special elections, as I think you probably should (you shouldn’t count both of Joe Manchin’s wins in this time period), the margin shrinks to 2.1 points (98.7 million votes for Democrats, 94.6 million for Republicans.
2. This is mostly attributable to Senate elections being held in different years.
This more modest discrepancy isn’t attributable to the representation of small states in the Senate. Instead, it is once again attributable to the fact that Senate races are held over the course of three elections, and Democrats performed best in the highest-turnout election.
Look at it this way: Suppose we take the popular vote wins for the parties (53 percent to 47 percent Republican in 2010, 53.7 percent to 46.3 percent Republican in 2014, and 56.1 percent to 43.9 percent Democrat in 2012), and assign the same number of votes cast to each election (it doesn’t matter how many). It works out to a narrow Republican popular vote win (.0042 points, to be precise).
3. This isn’t the small state bias at work.
There is still a discrepancy here -- 54 percent of the Senate seats for 50 percent of the popular vote -- but at least the arrow is pointing the right way now. Fifty-thousand votes changed across four Senate elections -- Alaska, Colorado and North Carolina in 2014 and Nevada in 2012 -- would bring the overall numbers into accord (of course, 28,000 votes switched across Montana, Virginia, New Hampshire and North Dakota would double the discrepancy).
But once again, it isn’t at all clear that this discrepancy is a function of the allocation of seats in the Senate. We have something of a controlled experiment in the House, whose seats are allocated by population. In the 2012 and 2014 elections, Republicans received more seats than their popular vote wins would have entitled them to. In fact, they received 3.8 and 4.4 percent more seats than their national margins suggested they should have won, almost identical to the disparity we see in the Senate. This isn’t a function of gerrymandering, this is about the nature of the Democratic coalition and how its voters are spread out.
4. We can’t compare the popular vote to House or Senate seats won.
All of this, however, is secondary to the largest problem with Vox’s analysis, and analyses like it. You simply can’t compare seats won (or electoral votes won) to national vote shares unless the numbers are wildly out of synch. This is because the parties aren’t trying to win national elections.
Consider some basic problems with the numbers above: How do we deal with the fact that Democrats didn’t field candidates in South Dakota in 2010? What about the fact that they technically didn’t field candidates for Kansas Senate in 2014, or Vermont Senate in 2012? What do we do with votes cast for Lisa Murkowski, Angus King, Charlie Crist (in 2010) and Larry Pressler?
True third parties also change the popular vote calculus. As I’ve noted, voters technically didn’t vote for a Democratic House in 2012. Instead, they voted for a coalition between either Democrats or Republicans and a sufficient number of third party candidates to achieve a majority. Because we have a winner-take-all system of elections (often referred to as “first past the post”), these votes are usually irrelevant to the outcome. Yet we have no idea what their preference would have been in a two-person race (or in a true proportional representation race).
Retirements also affect matters. If a party had a large number of retirements, its popular vote total will suffer. Democrats have had 16 retirements to Republicans’ 12, so this might have impacted Democrats’ vote totals at the margins. Likewise, if a party starts out at a substantial disadvantage, as Republicans did before the 2010 elections, they will have a difficult time making progress in the popular vote, because they’ll have to plow through a disproportionate number of the other side’s incumbents.
All of this alludes to the real problem here: In American first-past-the-post elections, candidates matter in a way they don’t in a proportional representational elections (the popular vote shares). Does anyone really think the votes that Ben Nelson won in 1996, 2000 and 2006 were the same as what we’d see if “generic Democrat” were on the ballot in Nebraska? Of course not. Likewise, Susan Collins’ vote totals in Maine don’t tell us much about the partisan preferences of Maine.
In other words, we can’t readily go from [candidate] to [party]. Yet this is exactly the thinking that these sorts of analyses invite. Democrats didn’t win 50 percent of the votes for the Senate (weighted by overall election turnout); Democratic candidates did. This is an important distinction.
This is especially true when you consider that unusually good candidates or unusually poor candidates can skew the overall results. Candidate quality matters. This is the sort of thing that tends to come out in the wash over time, but the weak showings for Todd Akin, Richard Mourdock, and Christine O’Donnell don’t help the Republicans’ national vote totals. Strong showings by high-quality Democratic candidates in Georgia in 2014 or in Nebraska in 2012 tell us about the number of seats “Democrats” should have won. Finally, if we had a national vote, wouldn’t we expect a better Republican turnout effort in Texas, or a better Democratic push in Rhode Island? Would these efforts cancel out? Probably, but not entirely.
This is why I say that as long as seats and vote shares are roughly in line, these discrepancies shouldn’t bother us much. They benefited Democrats through the 1990s, and seem to be benefiting Republicans today.