Improving the (Democratic) Primary

Part One: Primary Vote Allocation:

Primaries, like the general election, involve a public vote that allocates a slate of delegates to vote for a candidate. For the purposes of this discussion we will only consider the Democratic strategy for awarding state delegates:

Unpledged delegates, or Superdelegates, are distinguished party insiders (i.e. Governors, Congressional Representatives, Senators, DNC members, and distinguished party leaders). They represent about a sixth of all delegates. The others are computed in accordance with intermediate calculations. The first calculation we need to understand is the Allocation Factor:

The Allocation Factor is calculated using data from the past three Presidential elections (Jurisdictions without electoral votes are handled in a special way). Half of this factor is based on how many votes Democrats receive in a given state and the other half is based on how many electors a state is worth in the general election. The Allocation Factor is then multiplied by 3,200. Why 3,200? Because 2 is a party, 3 is a crowd, and 3,200 is a convention (Aside: There were about 4,700 total delegates at the 2016 Democratic National Convention).

The resulting number is the state’s Base Vote. This Base Vote can be increased (up to 35%) if a state decides to hold it’s primary later or hold it’s primary with other states. This helps explain why the Western states typically vote together (cluster bonus of +15% — the super tuesday south bloc votes too early to receive this bonus) and larger states, like California and New York, vote late in the primary cycle (late stage bonus of +20%). The Base Vote is used to calculate a state’s District, Statewide (At-Large), and Pledged delegates.

It is important to note how these delegates are expected to vote. District Delegates are awarded proportionally based on district votes. Statewide and Pledged Delegates are awarded proportionally based on the statewide votes. Pledged Delegates, like members of the Electoral College, are not legally bound to vote for their candidate. They can, and have, voted for others in the past. Although typically, those cases usually involve a candidate that has won delegates and subsequently dropped out (and issues a statement that releases them of their good faith obligations).

The Base Vote that we calculated above (and that is adjusted by bonuses) is then distributed between District and Statewide delegates. District Delegates account for 75% of the Base Vote and Statewide Delegates represent the remaining 25%. Pledged Delegates are additionally created to the tune of 15% of the Base Vote. In total, the standard amount of delegates is equal to 115% of the Base Vote. In summary, the equation used to determine a state’s overall number of delegates looks like this:

Let’s consider each number here:

3,200: This is necessary if there is a desire to represent a state’s vote with representative delegates. Since delegates are not given fractional voting power, which is completely possible, there exists a secondary effect of needing to round to the nearest whole integer at multiple stages during these calculations and during voting allocation.

0.5: This creates an equal distribution among two items, which are themselves unequal (the second term is, on average, about 1.5 times larger than the first). I would assume this is simply laziness. Assigning constants that appear equitable but that have no firm basis in reality nor describable intent is no way to craft a formula of such importance.

SDV/TDV: This provides representative weight based on the size of a state’s Democratic voting population. Given that the general election skews turnout towards swing states (by about 5–10 points) this seems like a somewhat appropriate way to provide representation. However, it does have a tendency to favor states with large (Democratic) populations, which doesn’t have any inherent advantages when trying to win the Presidency.

SAE/538: This factor weighs a state’s voting power in the general election. It is significantly larger than SDV/TDV, which is appropriate given the nature of the electoral college. It does not, however, take into account whether Democrats have a chance of actually winning those states.

1.15: This creates a separation between proportional representation based on district and proportional representation based on state. By including district-level representation and rounding, we have introduced another level of skewing representation that can be quite significant.

Bonus: The intent of bonuses is to encourage states to schedule their primaries in particular ways rather than requiring such action. This seems fundamentally equitable, but it would be more advantageous to craft a schedule with an intent of winning the general election, creating diversity in voter identity to avoid skewing future results, and/or being practical by conducting regional votes all at once. It is fundamentally irresponsible for a national party to defer to the state parties, who inherently act in their own interest, when creating a system with the intent of winning a national office.

Unpledged: These delegates enable conditions that subvert the will of everyday people voting. It is of some important to grant influence to those who might better understand which candidates would fare better in a general election and/or in office, but the fact that these delegates don’t vote with the rest of the population is problematic. It seems appropriate that the number of these delegates ought to be (dramatically) reduced and/or that the rules be modified with regard to when they vote. Or perhaps they should have a different role — such as moderating over (one of) the primary debates and/or posing questions to prospective candidates. Perhaps there should be a preliminary convention where the Superdelegates speak their thoughts and/or vote. Granted there are hundreds of them, but this could be worked in a way more conducive to their roles as party insiders.

An Alternative Solution:

Now that we’ve examined one of the current ways primary delegate allocation is handled — let’s consider a more prudent solution. Such a solution should be crafted in a way that is geared towards winning the general election. If we assume that winning an election is the primary reason for nominating a candidate then we ought to conclude that swing states > blue states > red states in terms of importance (switch blue and red for the Republican party). This calculus could change if voters in any given state use their power to abandon their winner-take-all strategy that they currently employ to allocate electoral votes. I digress. With this purpose and conclusion in mind it would make sense (for any party) to use the following formula in order to select a candidate:

In the first term, this formula considers each state‘s volatility by weighing the largest changes in voting margins against the closeness those margins. Since some states show extremely close average margins (i.e. Florida, Colorado, Ohio) and some states are won by such large margins (i.e. Massachusetts, Wyoming, Alaska) it is appropriate to bind the maximum adjustment to 3 and the minimum to 1/3 (3 representing the average deviation). The second term refers directly to the state’s electoral votes, which is the power a state actually has in selecting the President. This formula is adept at identifying the states that are most influential in any given Presidential election. It is therefore appropriate to give those states appropriate voting power in the primary selection. Note, this formula would not change the Democratic selection in 2016, but it would reduce the margin of victory.

If we consider the past four Presidential elections running up to 2016, the resulting offset would look like this:

The states colored purple have voted at least once for each party in each of the past five elections. Note that Missouri and Oregon have both had elections decided by less than one point during that time. It is also worth noting that the states that lose the most votes from this are California, Texas, and New York. The states that gain the most votes are Florida, Ohio, and Virginia. It is readily apparent how moving to such a formulation is advantageous to a party’s general election strategy. And while more comprehensive solutions might be considered, this is a substantial improvement over the ones that are currently employed.