While a single nation-wide primary would essentially require pre-existing universal name recognition, and an enormous amount of money for nation-wide media buys right away, staggered primaries allow candidates to build national recognition via the cheaper (and generally assumed to be more virtuous) art of small-scale, in-person "retail campaigning".
- The same, highly non-representative places, always get an undue influence on the process
- Caucuses that are insanely inaccessible, require several hours of your time, and are basically impossible to participate in for many (most?) people
- Violates one person one vote
- Convoluted delegate math makes for very complicated strategic decisionmaking
- Usual first-past-the-post problems (voting for your favorite candidate might mean "throwing your vote away" if they aren't one of the two most popular)
- It takes forever
Split the USA into 100 clusters of 4 or 5 congressional districts or equivalent (where DC is equivalent to 1 congressional district, the Pacific island territories are, combined, equivalent to 1 congressional district, and Puerto Rico and the US Virgin Islands are clustered together to be equivalent to 4 or 5 congressional districts). Prioritize contiguity of the clusters first, but also try to equalize population by putting more lower-population congressional districts or equivalent in clusters of 5, and more of the higher-population ones in clusters of 4.
Compute the population-weighted population density of each cluster (i.e. the average population density of a small area around each resident). Divide the set of all clusters into 4 equal sized tranches based on this density, representing roughly "urban", "high-density suburban", "low-density suburban", and "rural". Call these "density tranches".
If possible, divide the set of all clusters into 4 contiguous geographic regions, each with 25 clusters, such that each region has 5 clusters from each density tranche. If this isn't actually possible for some particular clustering, it seems like it probably should be if we allow ourselves to re-cluster. Bonus points if these regions come close to lining up with existing regional identities like "Midwest" or "Northeast" or "South".
We will choose four of these clusters, each of which will have an early primary. All four will be in different regions. All four will be in different density tranches. Each early primary will be three weeks apart. Then six weeks after the last early primary there will be a national primary.
No cluster can have an early primary if it has had one in any of the past four presidential election cycles (possibly excluding cases where an incumbent president of the party won the nomination for their re-election). Note that because there are five clusters for each (region, density tranche) combination, we can always find one that didn't have an early primary in any of the past four cycles.
Each primary will take place over the weekend, with voting open on both Saturday and Sunday, from 5AM to 11PM. Results will be released on Wednesday at noon. Partial results will not be reported.
Campaigning for each early primary will be limited to the two weeks before it. This will be enforced not by any rule, but by the fact that the location of each early primary will not be decided or announced until two weeks before it happens. So two weeks before the first primary weekend, a random cluster (excluding the sixteen clusters that have had early primaries in the past four competitive cycles) will be chosen as the location for the first primary. Then, a few days after the results of the first primary are announced, two weeks before the second primary weekend, a random cluster will be chosen (after eliminating from eligibility all clusters in the same region or the same density tranche as the first, as well as the fifteen recent early primary sites) as the location for the second primary. And so on. (Note: after the third location is chosen, there will only be five options for the last one as we will already know, by process of elimination, that it is necessarily part of a specific region and density tranche. There may even be fewer options if some of the five have had early primaries in recent cycles. This seems OK.)
The nominee will be selected by national ranked-choice voting in the national primary (which will be a true national primary, including people who already voted in early primaries).
The reason the early primaries matter at all will be...
There will be no official debates until after all early primaries are complete. There will be 3 official debates, happening ~5 weeks before, ~3 weeks before and ~1 week before the national primary. As suggested by Jamelle Bouie in this Twitter thread, these should be run by PBS, NPR, ProPublica, or other public broadcasting networks or non-profit public interest news organizations, rather than cable news networks or other for-profit news organizations.
Eligibility for the official debates will be determined solely by performance in the early primaries. A candidate will be eligible if any of the following is true
- In at least one early primary they either
- received the most votes, or
- received at least 20% of the vote
- In at least two early primaries they either
- received the second most votes, or
- received at least 15% of the vote
- In at least three early primaries they received at least 10% of the vote
- In all four early primaries they received at least 7% of the vote
- How do we deal with redistricting?
- Is the timescale too compressed for absentee or mail-in voting? To what extent do we even want that, if the point of the early primaries is in-person retail politics? If you're out of town for the whole campaigning period, we probably don't want you to vote. (You'll still get to vote in the national primary, which actually determines the nominee.) But we still want the accessibility gains of no excuse mail-in voting. Maybe have no excuse mail-in voting, but you need to receive the blank mail-in ballot at an address in or near the district? The two week window may still be infeasible though...