The Michigan Democratic Party operates in direct violation of its own Rules and its own Directive on Proportional Voting (DPV). MDP’s process for slate voting, and the example it provides to members, promotes a mathematical formula for calculating seats the DPV itself repeatedly asserts is the wrong formula. A well established body of research (e.g., Lijphart, Benoit, & Gallagher) shows that the formula used by the MDP is inherently less proportional than the procedure mandated by MDP Rules 2.A.8, Article 11, and the DPV itself. The formula promoted by the MDP is mathematically biased, empowering large groups over small groups. Worse, the larger the size difference between two groups, the more heavily the MDP formula favors the larger group over the smaller.
In principle and practical effect this is identical to gerrymandering. The people in charge of running the elections have chosen an election method that systematically disenfranchises some and advantages others.
For example, if the elections in CD7 and CD14 are typical examples, as many as 56 (possibly more) of the people “elected” to the MDP State Central Committee — 16.3% — were not elected by the will of the people, but by a mathematical formula impermissible under MDP Rules. There is every reason to believe the CD7 and CD14 elections are typical examples.
Can Michigan Democrats effectively fight Republican gerrymandering, when the MDP itself uses a system that is identical in principle and practical effect?
One way to see this flaw in the MDP’s formula is to notice that it allows slates with significantly less than 100% of the vote to nonetheless win 100% of the seats. For example, in an election with five seats to fill, it allows a slate with less than 84% of the vote to win 100% of the seats.
As I wrote in my first post on this blog, neither I nor Michigan for Revolution has the slightest interest in a weak MDP. We all want a strong MDP.
This hypocrisy on proportional representation undermines the Party’s strength on core democratic values the Party claims to champion.
So let’s fix it. Now.
Fixing it now makes a bold progressive statement to voters across the state. The Michigan Democratic Party doesn’t fool around when it finds a problem, they proactively fix the problem. Even when it may be against the interests of powerful factions, the MDP doesn’t find excuses to hide behind or defend mistakes, the MDP does the right thing on principle.
Democrats do the right thing on principle.
This is an opportunity for Democrats to stand up and correct our own Party’s gerrymandering problem, as we’re standing up across the state to correct the gerrymandering problem in our state government. If Democrats aren’t willing to fix the gerrymandering problem in our own Party, where Democrats have all the power to correct the problem, what credibility can we claim on anti-gerrymandering election reform?
Below I’ve provided an overview of the relevant rules and several simplified examples that illustrate the flaws noted above. At the end of the post there’s a downloadable PDF that contains a step-by-step walk through of a slate voting election using real world-data from the CD14 Caucus election this past February. Rather than the simplified version, in the PDF I use the actual system in every detail to demonstrate the flaws in the MDP’s method, including where it directly conflicts with the explicit instructions in the DPV.
Lastly, at the end of the PDF, I’ve provided a link to a short document which provides a simple step-by-step process that could replace the current explanation of slate voting in the DPV. I put that forward purely as a suggestions for consideration and discussion. If anyone has suggestions for improvements, or a different idea, please let me know. I’ll be happy to publish it here where we can discuss it.
If anyone would like to send me the raw vote data from any CD caucus from this past February’s election, I’ll be happy do the math and publish a complete document showing exactly how many seats on the SCC from that CD were filled by this mathematical gerrymandering against the will of the people. I’ll show every step in such a way that anyone with basic math skills can verify my results. If I get the data from every CD, I can compile it to show the total effect on the SCC across all CDs.
Of course, anyone can do the calculations. I encourage anyone skeptical to do the math themselves. I’ll be happy to meet with anyone interested in the details and work through them together, and of course I’ll answer any questions in the comments here.
This is an opportunity to make the MDP stronger. Let’s build a stronger MDP together.
Overview of Relevant MDP Rules and Simplified Examples
Michigan Democratic Party Rules explicitly require proportional voting at two different steps. MDP Rule 2.A.8 reads in full,
Proportional voting shall be used in the election of delegates and alternates to any County or Congressional District Convention, delegates and alternates of the Democratic State Central Committee, and members of any County or District Executive Committee.
MDP Rules Article 11 (paragraph 3) reads in relevant part,
All State Convention Delegates shall cast their share of the vote allocated to their County or District by the Convention Call using proportional voting as set forth in the Directive on Proportional Voting.
Let’s unpack this a little bit. Every two years, the MDP allocates to each portion of a County within each Congressional District (CD) a specific number of votes according to a formula found in the MDP Rules. When each CD Caucus convenes at the MDP Spring Convention to elect State Central Committee (SCC) Delegates and Alternates, the number of votes allocated to each county is divided exactly equally among all credentialed voters present at the Caucus. That’s what Article 11 is talking about where it reads “All State Convention Delegates shall cast their share of the vote allocated to their County or District by the Convention Call…“ (emphasis added).
Article 11 goes on to say that the voters present in each CD Caucus shall cast their share of votes “using proportional voting as set forth in the Directive on Proportional Voting”. It’s important to be clear that the MDP Rules require proportional voting at both steps. Proportionality by county, and proportionality by voter support.
Proportionality by county is exact. In each CD, each voter from a given county has exactly the same number of votes as each other voter from that county. Per the DPV, if in a given CD there are 100 votes allocated to a given county, and there are 50 properly credentialed voters from that county at the Caucus, each voter receives exactly 2 votes. If there are 200 voters from that county, each voter receives exactly 0.5 votes. Fractional votes are allowed.
Proportionality by Voter Support
Under the same MDP Rules that require proportionality by county to be exact, proportionality by voter support should also be exact, or as close as mathematically possible. Each slate should win a percentage of seats exactly equal to the percentage of the vote the slate received. If the slate received 40% of the vote, it should receive 40% of the seats. For example, suppose Slate A receives 40% of the vote and Slate B receives 60% of the vote. Then,
Slate A 40% x 5 Seats = 0.40 x 5 = 2.00 Seats
Slate B 60% x 5 Seats = 0.60 x 5 = 3.00 Seats
Slate A wins 2 seats, Slate B wins 3 seats.
Calculating Seats Won in a Slate Voting Election
Of course, the multiplication will not usually result in just a whole number, so we need a method to resolve elections when we get decimals. Here is the procedure.
- Calculate each slate’s percentage of the total vote in the election.
- Multiply each slate’s percentage of the vote by the number of seats to be filled.
- Award each slate a number of seats equal to the whole number part of the answers.
- If there are any unfilled seats after Step 3, compare the decimal part of the answers, and award the next unfilled seat to the slate with the highest decimal.
- If there are still unfilled seats, repeat step 4 until all seats have been filled.
- If there is a tie, use a tie breaking procedure (we’ll return to ties at the end of the PDF below).
In the above example of an election with five seats to fill, instead of 40% to 60%, suppose Step 1 revealed that the vote was Slate A with 33%, and Slate B with 67%. Then,
Slate A 33% x 5 Seats = 0.33 x 5 = 1.65 Seats
Slate B 67% x 5 Seats = 0.67 x 5 = 3.35 Seats
Slate A won 1.65 Seats, the whole number part is one. Slate A wins one Seat.
Slate B won 3.35 Seats, the whole number part is three. Slate B wins three Seats.
There is still one seat unfilled. Slate A’s decimal is 0.65. Slate B’s decimal is 0.35. Since 0.65 is greater than 0.35, Slate A wins the next unfilled seat.
There are no more unfilled seats.
Final Results: Slate A wins a total of 2 Seats. Slate B wins a total of 3 Seats.
The above is the simplest method that meets the requirements of the MDP Rules. This method is also as close to exactly proportional as mathematically possible. This means, the above method (or a mathematically equivalent one) is the only method that MDP Rules allow. To allow any other would, at minimum, require providing a reason for using an inherently less proportional method in the face of MDP Rules requiring proportional voting.
It’s possible to imagine that a less proportional method might have some advantage over the proportional method outlined above. And if that advantage were clearly called for by some MDP Rule, there would be a legitimate question of balancing the importance of the two rules. If there is no rule supporting the less proportional method, there is no justification for using a less proportional method when the most proportional method is readily at hand.
There are no MDP Rules that lends the slightest support to any method less than maximally proportional.
The MDP’s Disproportional Method
Without justification, the MDP wants everyone to use a similar but critically different procedure. In Step 2, instead of multiplying by the actual number of seats to be filled, the MDP wants everyone to multiply by the number of seats to be filled plus one.
Running the above example using the MDP’s method,
Slate A 33% x 6 Seats = 0.33 x 6 = 1.98 Seats
Slate B 67% x 6 Seats = 0.67 x 6 = 4.02 Seats
Slate A won 1.98 Seats, the whole number part is one. Slate A wins one Seat.
Slate B won 4.02 Seats, the whole number part is four. Slate B wins four Seats.
There are no more seats to fill.
Final Result: Slate A wins a total of 1 Seat. Slate B wins a total of 4 Seats.
Note that with the MDP’s method, both slates increase their number. Under the proportional method, Slate A’s number was 1.65. Under the MDP’s method Slate A’s number increased to 1.98. An increase of 0.33.
However, Slate B’s number increased much more. From 3.35 under the proportional method, to 4.02 under the MDP’s method. An increase of 0.67. Using the MDP’s method, Slate B’s number increase more than twice as much as Slate A’s number. That extra inflation took Slate B over the the line and into the next higher whole number, securing the next seat.
This is what I mean when I say that the MDP’s method disproportionately advantages larger groups over smaller groups. It artificially inflates every group’s numbers, but it always inflates a larger group’s numbers more than a smaller group’s numbers. And the larger the size difference, the more heavily the MDP’s method favors the larger group. In the CD14 election this past February, the MDP’s method inflated the largest group’s number by 868% more than it inflated the smallest group’s number, and 141% more than it inflated the middle group’s number.
Without justification, the MDP is insisting on a method that produces results that differs from the most proportional procedure in no other way than to disproportionately and systematically reduce representation of smaller groups and disproportionately increase representation of larger groups. The MDP’s method always favors larger groups over smaller groups. There are formal mathematical proofs that establish this fact, such as in Gallagher (page 495).
Here’s another way to think about the differences between the two methods. With five seats to be filled, the proportional method says each slate should win one seat for every full 20% of the vote it received (100%/5 seats = 20% per seat). In the example above, where Slate A has 33% of the vote, we can say it has won one seat and is 7% short of 40%, where it would win another seat. Similarly, where Slate B has 67% of the vote, we can say it has won three seats and is 13% short 80% where it would win another seat. Since Slate A is closer to winning the next seat than Slate B, awarding the seat to Slate A is mathematically the closest we can come to exactly proportional. Awarding it to slate B is disproportional by 13%, where as awarding it to Slate A is only disproportional by 7%. In this example, which is typical, the MDP’s method is 6% more disproportional than the maximally proportional method.
Here’s a chart of example elections that illustrate this point. Each row is an example election with five seats to fill and two slates running, showing the results under the proportional system and under the MDP’s disproportional system.
At 30 / 70, Slate A has won one seat and is halfway to the next; Slate B has won 3 seats and is also halfway to the next. Under the proportional system, this is a tie, as is fair (I explain tie breaking procedures in the last section of the PDF below). The MDP’s disproportional system doesn’t produce a tie, it just awards the next seat to the larger group.
At 31 / 69, Slate A is 9% from the next seat, while Slate B is 11% away. The MDP method still awards the seat to the larger group. Similarly for 32 / 68, and 33 / 67. The MDP’s disproportional system produces a similar pattern every time two (or more) slates are objectively equally distant from the next seat. In the examples above, under the MDP’s method the smaller group has to win 4% more votes than the larger group in order to win that next seat. That is not proportional. This is putting a thumb – and three fingers – on the scale to tip the balance towards larger groups.
These effects pile up. Consider – a slate with much less than 100% of the vote can nonetheless win 100% of the seats. Consider this example,
Slate A 16.665% x 6 Seats = 0.16665 x 6 = 0.9999 Seats
Slate B 83.335% x 6 Seats = 0.83335 x 6 = 5.0001 Seats
Slate B wins all five seats.
In this example, under the MDP’s system, less than 84% of the vote is enough to win 100% of the seats. Under the exactly proportional system, where we multiple by the actual number of seats to be filled,
Slate A 16.665% x 5 Seats = 0.16665 x 5 = 0.83325 Seats
Slate B 83.335% x 5 Seats = 0.83335 x 5 = 4.16675 Seats
Slate B wins four seats, and Slate A wins one seat (because 0.83325 > 0.16675).
Using the MDP’s method, this happens all the time.
A system that awards 100% of the seats to a slate with less than 84% of the vote is prima facie not proportional. A system that pushes a larger group past the finish line when a smaller group is closer to the finish discriminates against groups in the minority. In every feature, the MDP’s method degrades core democratic values of fairness, proportional representation, and minority representation and empowerment. To claim this method is acceptable under MDP rules is an assault on these core democratic values, and the MDP Rules themselves.
Let’s fix this. Let’s fix this now.
(The above is a simplified version of real-world problems. In the attached PDF I walk through the entire slate voting process step by step, using real-world data to illustrate the issues discussed above.)
Sarah Horn assisted with this post.