More on the Myth of Democrats' 20-Million-Vote Majority
Dylan Matthews has replied to my response to his piece last weekend, which discussed the fact that current Democratic senators won 20 million more votes than their Republican counterparts. There are a few issues that require further discussion and clarification.
As a housekeeping matter: We’re discussing the total number of votes won by victorious Democratic senators vs. Republican senators, as well as the vote shares won by all candidates of both parties (what we commonly refer to as the popular vote). These are mouthfuls, so for simplicity’s sake, I’ll refer to them as “total votes” and “vote shares.” When I talk about our system of elections -- voting for individual candidates in states or districts -- I’ll call it “first past the post,” or FPP. When I refer to voting for parties, I’ll call it proportional representation, or PR. Note that there are forms of proportional representation that do involve voting for candidates, but I’m not sure that’s really what we’re discussing here.
There are two “big picture” things to address up front. First, contra Matthews’ contention, I do think that we disagree on the numbers, or at least the way they were presented. He cites Nathan Nicholson at FairVote for the total-vote statistic: that Democratic senators received 20 million more votes than Republican senators. He then reproduces a chart that he claims illustrates this phenomenon. It simply doesn’t (as he notes in his follow-up piece). It instead shows the vote share across four elections.
Matthews then goes on to note that Democrats received more votes overall than their vote share suggested (as displayed in the chart) in 2008 and 2012. It’s clear there’s a shift in the metric being discussed here, because this claim isn’t true using the original total-vote metric: In 2012, victorious Democratic senators won 82 percent of the vote but 76 percent of the seats.
Matthews finally draws conclusions -- that this is about the “small state bias” of the Senate, that the Senate is undemocratic and should be abolished -- that I don’t think really flow naturally from the total-vote metric (more on this later), but that arguably would flow from the vote-share metric.
Taken as a whole, this was misleading. Judging by the reactions that his piece and my response got, more than a few people were left confused about what the “20 million” statistic actually represented, especially given that he only discusses vote share in his concluding paragraph (for more along these lines, see Philip Bump here). This was the main thrust of my article.
Second, I should have been clearer when I used the term “misleading” in my piece that I didn’t mean Matthews was being intentionally misleading. I’ve followed him in the past and have found him to be intellectually honest and well worth reading, and didn’t mean to imply he was dishonest here. I suspect that this was just a short “thought piece” written in a hurry, or that something was left on the editor’s floor that might have clarified things further.
As for Matthews’ follow-up, it’s important to keep in mind what I was responding to, as the sweep of this discussion has grown considerably. Aside from the use of numbers, I did take some issue with his argument. As I understood it, it ran this way: (1) victorious Democrats won more votes than victorious Republicans; (2) this is an illustration of small-state bias; (3) the Senate is unrepresentative and should be abolished.
If claims (2) and (3) had instead been that, say, (1) illustrates the discrepancies that can come from midterm drop-off, and that we should therefore adopt compulsory voting laws (an argument he presents in his follow-up), I doubt that I would have responded because I don’t have strong feelings about compulsory voting laws, and at the very least this somewhat follows.
But I don’t see much support for the notion that (1) is because of (2). There’s also a broader issue here, which is that you can’t really use PR metrics (national votes for parties) to measure the fairness of FPP systems.
On the specifics: Matthews wonders why we should adjust for turnout, because Democrats’ strong 2012 was in part due to the nature of that electorate. If the midterm/general election discrepancy really were mostly about depressed turnout rates among Democratic groups, I’d be somewhat more sympathetic to his viewpoint. But it isn’t. It’s mostly a function of the fact that President Obama happened to be broadly unpopular among adults in late 2010 and 2014 (both of which were lower-turnout elections) and modestly popular in late 2012 for the higher-turnout election. If Obama’s job approval is 42 percent among adults going into 2016, these numbers are probably going to look a lot different.
Understand this: Arguing that we ought not adjust for turnout requires us to accept that Claire McCaskill’s vote total is somehow directly comparable to John Ashcroft’s much lower vote total (notwithstanding their identical percentages), or that Heidi Heitkamp’s narrow win in North Dakota is more or less the same as John Hoeven’s 50-point blowout. It seems highly unlikely that these sorts of discrepancies are mostly attributable to demographic changes in the electorate, rather than to candidate quality or generalized turnout issues. At best, the difficulty in sorting out persuasion vs. turnout is yet another reason why we shouldn’t use national vote totals across multiple elections as a metric for FPP fairness, though it might support an argument for compulsory voting or abolition of midterm elections.
Matthews writes, “Trende’s third subsection is entitled, ‘This isn’t the small state bias at work.’ This is odd to me, since, as he notes in that very subsection, this is being driven by facts of political geography that make the Senate’s small state bias very relevant.” The point is rather straightforward: The House doesn’t have a small state bias (to be more precise, it is very minor), yet it routinely exhibits discrepancies of the same magnitude that we see in the Senate. This is true even if we set aside the gerrymandering issue: Compare 2008, when Democrats won 55.5 percent of the two-party vote but 59 percent of the seats, to 2010, when Republicans won 53.5 percent of the vote and 55.6 percent of the seats.
Finally, Matthews notes that we can use national vote totals to evaluate the efficacy or fairness of FPP systems because “while candidates cast their votes for candidates, they get parties.” Sens. Susan Collins and Joe Manchin may beg to differ; I also suspect Democrats saw this differently during the health care debate, when the type of Democratic candidate needed to win in, say, Arkansas, meant that there would be no public option in Obamacare (to say nothing of a single-payer system).
But this is a red herring. People cast their ballots for candidates, within states or districts, and not (directly) for parties, whether you’re a fan of proportional representation or not. A weak Republican in a strongly Republican state (like, say, Missouri, Indiana, North Dakota, Montana) or a strong Republican candidate in a strongly Democratic state (like, say, Maine) can cause parties to run up undersized or outsized margins in states in a way that wouldn’t show up in a pure proportional representation state. The success or failure of Democratic recruitment efforts in a single state in 2016 -- Florida -- could make a difference of a half-million or so votes in the national popular tally. We’d expect this to come out in the wash, as both parties will have their successes or flops, but not precisely. This is exactly what happens.
One can ultimately opt to look at the 375,000 votes that Ben Nelson received in Nebraska in 2006 and find some sort of meaning about nationwide preferences for a Democratic Senate, but I don’t think there’s a strong basis for doing so.
Of course, the much larger issue here is that we can’t look at national vote totals because the parties aren’t running national campaigns. Democrats tried to squeeze every last Democratic vote out of Arkansas rather than try to run up the score in New York; likewise with Republicans in Illinois and Alabama in 2010. Would the outcome look similar in a PR system? Yes. Identical? No. Do we have any idea how it would differ? Not really.
After all of this, one is still left wondering why, even as a matter of political philosophy, we’d care about vote totals, when the vote-share metric is readily available and more directly applicable to what advocates of PR (or abolition of the Senate) are really worried about. It isn’t really a surprising statistic, given what we know about turnout and which year Senate Democrats performed well in over the past few cycles.
Nor is it particularly relevant. Imagine, if you will, a five-person Senate, where 100,000 votes are cast for each seat. Democrats win, 51-49, in three of the states, and lose, 49-51, in the fourth. In the fifth state, an utterly unacceptable candidate wins the Republican nomination, and the Democrat prevails, 80-20. Democrats have 80 percent of the seats in the Senate, despite winning only 56 percent of the vote. There are perfectly valid ways to defend that outcome, but would anyone defend it on the basis that victorious Democrats won about four times as many votes as victorious Republicans? If not, why would anyone talk about this metric?