Really, this is unjustifiable. I used to believe the Democratic Party actually used “proportional representation” for its presidential nominating delegate allocation. But this is not even close.
“Realigned vote” is the shares after participants in candidate-supporting groups that were below 15% in any given precinct shift to their second choice. This is the only good part of the whole process. The “realignment advantage” is just an indicator of how much a candidate gains (or loses) at that stage: the realigned vote share divided by the initial vote share.
How you get from there to the final delegate allocation is clearly some sort of black magic. It goes by way of this weird phantom called “state delegate equivalents”.
And somehow out comes an allocation in which a candidate who started off in second place, even after realignment, not only comes in first in delegates but somehow is over-represented to about the same degree as the first place party in your average FPTP election. (Delegate advantage is the delegate share divided by the realigned vote share.)
This is not proportional representation. Whatever it is, it is no way to run a presidential nominating process.
And I have not even mention the great app failure, or the possibility that some precinct-level allocations were just calculated wrong. (See spreadsheet of errors.)
Sources for data: NYT, WaPo, IDP.
See also a breakdown of available delegates by county and precinct. (The IDP link above has it even more detailed.)
Fruits and Votes 
It is a system that clearly should die and be replaced.
LikeLike
No, not very proportional, but definitely not winner-take all either. My first guess is that the conversion from “SDEs” to national delegates gives more weight to small precincts, and/or precincts in small counties, than to large precincts. And my second guess is that this is by design, not by accident. In other words, the arrangement is loosely analogous to the way the Electoral College gives each state two votes for its two senators, regardless of population. And gives even the smallest state one vote for its representative in the House, even when it’s population is less than 1/435 of the national population.
LikeLike
That was my thought the other day when the count was ‘nearly finished’. It appeared to me that Bernie may have won many precincts by wide margins but got no ‘bonus’ for the added margins, while Buttigieg may have won more precincts by slim margins (easy enough to do in a race with 8 contestants). The county-won totals give the edge to Mayor Pete but these are lightly populated areas, it appears. The awful and partisan and damned incompetent Iowa Democrats did the rest?
LikeLike
And the good news is that the Nevada Democratic Party is in a complete mess trying to put together a caucus under almost identical rules after abandoning an app that is now radioactive. I laughed out loud at a fairly damning article that suggested Acronym named their software house ‘Shadow Inc’ because ‘Moustache Twirling Globalists LLC’ was unavailable.
LikeLike
Iowa and the Democratic Party really have to ask themselves after this catastrophe whether the tradition of the caucus and being ahead of New Hampshire is really worth it anymore. Primaries are more accessible, easier to administer, and don’t necessitate any black box SDE shenanigans to account for differential turnout. Simple statewide delegate allocation with a reasonable threshold is better for everyone.
LikeLike
It’s an MMM system with some malapportionment issues in the both tiers.
Each Congressional District has a number of delegates (I believe based on a combo of the 2016 presidential and 2018 gubernatorial results). CD 1 and 2 each have 7 national delegates, CD 3 has 8, and CD 4 has 5. The state delegates from within that congressional district at the state convention pick the nominees so it’s proportional based on state delegates won IIRC. The number of state delegates isn’t exactly proportional due to rounding (and due to the fact that smaller rural precincts are more likely to have less than 7 delegates, and therefore a higher threshold than 15%, so this benefits candidates who do disproportionately well in rural areas).
The fact that CD4 is small and rural has the same effect when it comes to picking national delegates. Buttigieg won 2 delegates there, Bernie 1, and Biden and Klobuchar 1 each. This is on a split of roughly 108 state delegates for Buttigieg to 100 for Bernie.
Then there are 14 national delegates, again determined by how many state delegates each candidate has. Despite getting fewer votes, Buttigieg had more state delegates total, so he won more of those seats despite getting fewer votes.
http://www.thegreenpapers.com/P20/IA-D#0203 has the math.
This is of course a travesty. If the state based system is to be retained (I believe it should be abolished), there must be a proportional allocation at the state level. I’d prefer there be an AV-style ballot, with anyone who fails to cross the threshold (which I believe should be somewhat lower than 15%) distributed to next preferences.
LikeLike
15% is an arbitrary number that dates from the McGovern commission, which saw it as a way of protecting minorities in a state party. A far better threshold would be Droop—Integer(Votes/Seats+1)+1
LikeLiked by 1 person
Is “Integer” there floor (rounded down) or ceiling (rounded up)?
LikeLike
Irish Electoral Act 1992:
The Droop quota is the smallest number that can elect the correct number of candidates. You do get edge conditions for the last vacancies because of exhausted votes.
LikeLike
I always viewed Droop* in simplified form as being Seats/Votes, rounded up the next whole number. 100 votes for three seats produces as quota of 34. 100 votes for four seats produces a quota of 21.
LikeLike
Mark
That is wrong. The Droop quota where V=100 and S=3 is (100/4) +1 =26. The Droop quota where V=100 and S=4 is (100/5) +1 =21. Your worked example where S=3 gives a far higher figure than the Droop quota. For S=4 your method would actually return 100/4=25.
Your method produces different results and is clearly not the Droop quota.
LikeLike
It’s not my method, it was my math. I meant to have divided 100 by 4 for three seats and by 5 for four seats.
Votes/Divided by Seats +1 with the quotient rounded up to the next whole number. At least that is what I meant to convey
LikeLike
But you have two separate multi-member tiers, both of which are “proportional” (the closest parallel I can think of is the Guatemalan system), so what quota do you use?
In CD-4, you’d have a quota of 16.67%, while in CD-3 you’d have 11.1%. I have an issue with the varying quotas because it benefits candidates who do better in rural areas and harms candidates who do well in urban areas (well, not intrinsically but it does in Iowa).
For the statewide tier you’d have a quota of 6.67%.
And none of that changes the big issue that rather than hearing who is viable at the statewide level and realigning based on that, each caucus site uses their own 15% calculation (or higher than 15% if they have 6 or fewer delegates for the next round), and rounds to the number of delegates they have. This again benefits candidates in smaller, rural areas (which are more likely to have 6 or fewer delegates) and hurts candidates in larger, urban areas (which are more likely to have 7+). Candidates who have disproportionate appeal in rural areas while still garnering 15%+ in rural areas will do very well on delegates, while the reverse is not true.
Elizabeth Warren got hurt by this most of all (she needed 89.925 final realignment votes per SDE received, compared to just 69.566 final alignment votes per SDE for Biden); Sanders also did relatively poorly at 81.237.
Unless each caucus were to phone in results and then wait to hear which candidates crossed the threshold statewide, then have a realignment/distribution of preferences for excluded candidates, count and award state delegates (allowing fractional delegates), then phone in results after that realignment to hear which candidates crossed the congressional district threshold and do the same thing. Given how hard it was to report results already, this seems impossible. It also would potentially exclude a candidate like Klobuchar who did well enough to earn a delegate in one congressional district but not statewide.
Klobuchar got saved by the late reporting of the result. If it had been reported on election day, she would have clearly seemed like an also ran and likely not done as well in New Hampshire.
Of course, all of this is just trying to put a bandaid on a broken system rather than doing something sensible like having a single national primary with AV.
LikeLike
In a separate thread I’d suggested statewide allocation and almost everyone had suggested abolishing caucuses. I’m not sure a single national primary would necessarily work very well. A staged primary with small states front-loaded and large states back-loaded strikes me as about right, but the order should be random and should change every cycle. Iowa and New Hampshire could perhaps be consoled with beauty contest primaries at the start of the season.
LikeLike
Pingback: The strategic voters’ nightmare that is US Democrats’ “proportional” system | Fruits and Votes
“… “Realigned vote” is the shares after participants in candidate-supporting groups that were below 15% in any given precinct shift to their second choice….”
The League of Woman Voters is opposing this, right? Because “it gives some voters two votes”, right?
LikeLike
The League is opposing it? Really?!? It seems to me that only one of those votes counts, so what’s their problem?
LikeLike
Being sarcastic. The LWV opposed IRO-AV in Alaska a decade ago because it “gave some voters a second vote”. This after a half-century of direct primaries in the US, without a peep of complaint from the League about those voters who, eg, in 2008 “dated Clinton, married Obama” electorally.
I don’t see any great difference in detail between primaries and IRO-AV on this point, especially as party-column primaries can easily shade into Californian “top-two” primaries (no one in America would say “California no longer holds primary elections”0, which is basically two-round runoff, which shares this feature with IRO-AV.
I’d answer that as “Yes, they do get two votes [assuming there are two rounds. if it’s full elimination then with, say, six candidates, you get five votes, and so on]. But then, everyone gets two votes [five votes, etc]. And if you support one of the Top Two (or occasionally – looking at you, France – the Top Three), then you get the luxury of casting all three of your votes, all three times, for your most favourite candidate: no need to compromise. The only ones who get to transfer their votes among two or more different candidates, at different times, are those who don’t get the luxury of staying with their most favoured candidate. So, to anyone mathematically literate, complaining that the minor players’ supporters got to vote for TWO DIFFERENT CANDIDATES!!! is like complaining that your college roommate had all the fun of going to ten different job interviews before they got hired, which you missed out on because you got an offer of employment on your very first interview.”
LikeLike