Factions by the Numbers

14 mins read
Photo by Tetiana SHYSHKINA | Unsplash

The 2021 convention provides us with an opportunity to examine dynamics of influence within the DSA national organization and the country.  Publicly-reported votes and data provided by the National Office provide us with points from which to perform a data-driven analysis of where (and which) factions hold support across our organization, including the geographic shape as well as the intensity of that support.  The 2019 ballots allow us to track trends over the last two years as well, and see how those support bases have changed over time.

This post will examine national faction-driven dynamics, particularly how they have shifted between the 2019 and 2021 conventions.  We will take a deeper dive on voting trends across resolutions and amendments in later posts.  

The purpose of this report is not, however, to establish or erode one faction’s legitimacy over another. We’ve tried to take a lot of care to get as accurate an assessment of factional alignment as the data will allow. Trying to fashion some hidden moral victory from voter preferences is irrelevant because there are no such things as moral victories, just victories. Moreover, the dynamics of leadership bodies, how they operate or vote, is not simply a function of voter intentions. In fact, voter intentions are more often than not wholly irrelevant to their day-to-day work. 

Our purpose is to try to get a map of factional alignment across the org and develop a better sense of how it is changing, as well as what this information can tell us about the often-mysterious internal lives of these groups. We want to see the directional movement of our organization in terms of ideology and influence, to better understand where we might be headed. 


To begin an analysis of trends, we ran an “alignment” analysis of each of the ballots cast in the 2021 NPC election. We focused on the top five rankings on each ballot to determine a primary alignment. We will discuss how each alignment was calculated below, but generally, we grouped the NPC candidates into “factions” based on discernable common preferences in votes with some degree of discernable coordination. These factions can then be further subdivided between “major” and “minor” factions, with major factions having their own candidates and organizing their votes to give their candidates the most favorable NPC possible, and minor factions simply choosing among the candidates on offer but still pushing their preferences in an organized way.  

In this piece, and likely in future pieces, we will use the term “faction” to refer to these organized blocs, instead of the perhaps more common term, “caucus.” The reason for this is because caucus has a very technical meaning within DSA and not all the groups we discuss here match that framing. Some are caucuses, some are slates, some are slates with separate caucuses as their principal supporters, some are magazines, some are separate organizations altogether. They are all, however, working to organize their votes to meet a specific outcome and we will refer to them all as “factions” so as to create as little confusion as possible. 

  • Bread & Roses (B&R), major faction
  • Class Unity (CU), minor faction
  • Green New Deal (GND), major faction
  • Libertarian Socialists (LSC), minor faction
  • Reform & Revolution (Re&R), minor faction
  • Renewal (RNW), major faction
  • Socialist Alternative (SALT), minor faction
  • Socialist Majority (SMC), major faction
  • Tempest (TMP), minor faction

We created a “generic ballot” for each faction and assigned each ballot an alignment based on whether they met certain ballot preference criteria.  Of the 1,146 ballots cast in the 2021 election, we were able to establish a factional alignment for 928, leaving 218 as “unaligned”. This analysis does not assume that individual voters necessarily had any formal relationship with these factions, or even purposeful adherence to their ideologies. It does show where there is influence and alignment in goals for the organization overall, or, at the very least, that a delegate found that faction’s vision most compelling. 

2021 Generic Ballots

Because there were many fewer candidates than usual for NPC elections, determining voting preferences across factions had some challenges. The unique nature of each faction’s preferences helped simplify this in some ways. SMC voting preferences were the easiest to track, as the only faction with a formal slate of five candidates. GND voters were similarly easy to track, having three candidates which were unique to their ballot. If a voter had three or more SMC or GND candidates in their top five, then they’re a SMC or GND voter and can’t be mistaken for anything else.

Bread & Roses posed some challenges. B&R ran a slate of three, and published a voting order. The generic B&R ballot also always had their three candidates at the top, then Matt Miller, then almost always a GND candidate. However the complication comes in the form of minor factions which share a lot of their preferences with B&R.

Three minor factions specifically, Reform & Revolution (Re&R), Tempest (TMP), and Socialist Alternative (SALT), all included the entire B&R slate in their top five. Re&R and SALT even included B&R’s fourth choice, Matt M, in their top five as well. However the specific nature of their ballots helped correct for this. 

Re&R and SALT both put Matt M above B&R, something a B&R voter would not do. Re&R, TMP, and SALT, also are identifiable by their remaining vote, which in all cases would be uncharacteristic of a B&R voter when comparing against the generic B&R ballot as a control. For Re&R this was Justin C, for TMP this was Jenbo, and for SALT they would simply bullet vote (meaning they only included candidates they strongly preferred on their ballot) Matt M first, then B&R, and then left the rest of their ballots blank. The fact that they voted as disciplined blocks, every time without deviation, also made it much easier to identify them. 

Renewal voters required some attention in order to discern because their generic ballot included only two unique candidates and then three independents who they had endorsed. People who ranked the Renewal candidates highly were very likely to rank Jenbo/Jen M./Justin highly, and vice versa. However, because the particular situation surrounding this convention may have skewed voting patterns, we erred against assigning a ballot to Renewal when alignment was unclear. There were 19 ballots that included the three independents in their top five but not either Kara & Austin; these ballots were counted as “not aligned” where no other factional preference can be determined. There were also 6 ballots which bullet voted for Kara & Austin but did not include the three independents in their top five, and they are considered “Renewal” voters for the purposes of determining alignment. 

The LSC voter was the most difficult to make out. They only had two candidates on their endorsement list that were unique to their ballot, Aaron W and Desiree JF. The other three they shared with Renewal voters. Unlike the problem with groups like Re&R and SALT versus B&R ballots, the LSC ballot also shared a lot of the same preferences overall as Renewal voters, meaning the presence of Aaron or Desiree somewhere on a Renewal aligned delegate’s ballot would not immediately identify them as LSC. Ultimately, we have to rely on intentionality to gauge the LSC voter, meaning they either ranked the two unique LSC endorsements in the top spots, or they ranked them over the unique Renewal candidates overall, meaning generally only the strong LSC voter would get detected. 

Collective Power Network (CPN) on the other hand was not able to be distinguished at all. Their official candidate endorsements included the entire Renewal ballot, the entire SMC ballot, and the entire GND slate. They also had no candidates or caucus members running to focus their support around to make them discernable as in 2019. We will discuss this in more detail in the portion on changes in voting trends between 2019 and 2021, but generally speaking their voters were effectively absorbed by other factions to make any level of clear organization among NPC ballots undetectable and any attempt to discern them entirely guesswork. 

2019 Generic Ballots

2019 was a very different race for a number of reasons. First there were 32 candidates running for NPC rather than 20 in 2021, making it in some ways much easier to discern delegates’ intentions when voting. This created far less overlap between distinct factions that would otherwise seem very similar. On the other hand, the nature of the slates running and the factions supporting them also posed its own problems.

Bread & Roses and SMC are relatively straightforward to interpret. Both were new caucuses in 2019, having only just formed months prior (though B&R had roots stretching further back). They both were definite factions running definite slates, making it relatively easy to track their preferences. They also shared very few  of their candidates with other factions. 

Tracking Build voters posed some unique issues. Build technically didn’t run a slate and instead only issued endorsements. There were seven Build members running for NPC who Build also endorsed. This is further complicated by LSC, who in 2019 had their own endorsements, and many, if not most, of their delegate members were also members of Build (including two candidates for NPC who dual carded with both factions). Fortunately however they had totally distinct generic ballots, though there was a great deal of sharing between them among their voters.

CPN in 2019 was in a very similar position as LSC in 2021. They, like Build, also technically did not run a slate for NPC and instead had a caucus member running as a candidate and also issued endorsements. Like Build and LSC voters, even though they didn’t have a formal slate, CPN voters voted in a very specific way that we can track in order to determine intention. Though like LSC in 2021, we are generally only able to identify the strong supporters, and any voters who liked only some of CPN’s choices but not others aren’t able to be identified reliably. 

2019 also had the San Francisco Slate, which ran three candidates, but was not run by a formal faction and none of its candidates were members of any caucus, similar to GND in 2021, though markedly there was a much greater degree of vote sharing between SF and other factions than with GND in 2021. There was also Momentum, a split from B&R, which is our only “minor faction” in 2019 since they had no candidates of their own but had their own clear voting order. 

It’s worth pointing out that SF, LSC, and Build had a great deal of overlap among voters. If their candidates had all run on a single slate, then an entire third of voters who are currently classed as “not aligned” could instead be classed under this larger slate. This would make this theoretical block account for 40% of voter preferences of the entire convention. This could be attributed to a coordinated strategy between the slates to share votes, or simply due to the candidates supporting many of the same proposals such as “Pass the Hat.” We can’t know for certain but it’s worth pointing out. 

Analysis of Factions 

Now that we have all the ballots from both 2019 and 2021 assigned an alignment, we can start looking at the numbers. 

Here we can see the breakdown of votes by the total convention. In 2021 the four major factions performed roughly the same in terms of vote totals, with only 73 votes separating the largest vote getter and the smallest among them. Comparing these numbers to 2019, we can see that B&R’s vote basically held steady, and even increased their footprint slightly, while SMC’s had a significant decrease. 

Here we see the breakdown by the strength of the factions ballots. A “strong” ballot is one where that delegate voted exactly the way the faction wanted them to for all five of their top spots in any particular order. A median ballot they shared four out of five spots, and weak three out of five. Anything less where no other alignment can be determined is considered a “not aligned” ballot. All minor factions ballots are basically by their nature only “strong” ballots because otherwise they are too difficult to detect.

B&R had far and away the most disciplined voters, even after separating out Re&R, Tempest, and SALT ballots which shared their three candidates all in common, and similarly voted in near lock step. There was however far more deviation among B&R voters than those other groups. 55% of B&R voters put a GND candidate as their 5th spot (usually this was Gustavo G. but not always), while 30% gave their 5th spot to an SMC candidate, usually Jose LL but, again, not always. 

Renewal and GND voters were in some ways the opposite of B&R voters in terms of disciplined voting, with only 14% and 10% of their total votes coming from disciplined voting respectively. When compared against B&R and SMC, two factions that have existed for at least two years, this likely reflects a lack of formal caucus infrastructure, relying instead on broad support and positive word of mouth for votes. It’s worth mentioning that Build and LSC had similar ratios of strong versus weak support in 2019, who both had far more established formal infrastructure, both having only 20% of their total votes coming from strong ballots. 

The only similarity these four factions share that might explain this is maybe the fact that none of these groups had a definite caucus attached to a definite slate of the same name, all instead were either simply slates without formal caucuses or caucuses without formal slates. Perhaps this made it more difficult to translate support from the caucus to support for a set of candidates when they aren’t immediately identifiable with one another to a delegate who is having to take in a lot of information very quickly during the course of a national convention. 

It’s also possible that this similarity in voting patterns is simply due to votes being cast against what is perceived as established tendencies in the organization rather than strong support for a definite program and that the nature of these sorts of votes are simply more pliable. These numbers alone don’t provide a clear answer as to the reasons a person cast a ballot in a certain way, and we will not try to figure out these questions here.

Analysis of Chapters

Raw vote totals from one convention can only paint some of the picture, especially if we want to know where factions support is coming from, not just how much of it there is. For this we can add shape to these votes by breaking them down by chapter. Because chapters tend to remain constant between conventions, they help give us a sense of where fractions are drawing their support, and also how factions are interacting with chapters on the ground level of our organization. 

These figures give us a breakdown of alignment based on chapter. Here, in order for a chapter to be considered “aligned” with a particular faction, that faction needs to be the highest vote getter and receive at least 30% of the votes. Ballots that are considered “not aligned” also compete in this comparison and only chapters with five or more voting delegates are mapped. 

The first figure shows us the vote breakdowns by chapter, while the second shows us how alignments changed in the aggregate. Chapters with a tie between two factions are counted for both, while a tie between three or more factions are counted as “not aligned.” The same standards were applied for both 2019 and 2021. Chapters for which there was no data in 2019 are included in the first figure but not in the second. 

Chapters with no alignment represented the largest block in 2019 but in 2021 reduced considerably. No chapters which were non-aligned in 2019 remained so between conventions. In fact all the non-aligned chapters in 2021 had some alignment in 2019. Renewal and B&R appear to have been the main preferences for previously unaligned chapters, with the rest going to GND or Class Unity. 

While Bread & Roses’ total share of ballots in 2021 dropped slightly from 2019, their total chapter alignment count grew from 11 to 14. They retained the majority of the chapters that aligned with them in 2019, the only faction to do so, and were able to make up any drop-off by picking up some previously unaligned chapters. Notably, the minor factions that voted with B&R for NPC built their base of voters in chapters where B&R formerly occupied the top spot. The single chapter where Tempest was the main alignment was a former B&R chapter, and the two chapters where Re&R were able to achieve a plurality of votes (though not enough to be the main alignment) were also B&R chapters in 2019. 

Renewal took a little bit off of everyone’s plate. They transferred chapters from every major and minor faction in 2019 to themselves except CPN. The largest set that Renewal picked up was previously non-aligned chapters moving over to Renewal, with former Build and SMC chapters about even for second place. GND had a similar process though by far pulled most of their votes from places where SMC were formerly strong. This isn’t only reflected in the second figure, but also looking at vote totals in large chapters like New York City and Los Angeles where SMC had previously won decisively.

Perhaps the most dramatic shift between 2019 and 2021 is the considerable shearing off of Socialist Majority’s base among chapters. Their total vote share of the convention went from 21% to just 13%, having them go from the largest major faction by total ballot alignment to the smallest. Their base among chapters went down by almost two thirds between conventions,  going from 14 in 2019 to only 5. They made almost no new ground either, holding onto 4 of their previous chapters and only picking up one new chapter from Build. 

A recent article in Socialist Forum attributed SMC’s performance to the method of voting used, Single Transferable Vote or STV. We aren’t going to wade into the remarkably contentious debate around STV versus Borda voting methods, but we will point out that based on the numbers, SMC’s decrease in NPC seats can’t be explained by voting methods alone. A much more plausible explanation is simply that their voting base collapsed in several chapters that were critical to their success in 2019. 

Three of the five biggest chapters in the country had gone comfortably for SMC in 2019-New York City, Los Angeles, and Boston-with the other two, Chicago and DC going to B&R and CPN respectively.

In NYC, SMC went from a 7 point lead in 2019 to losing to GND by 13 points. In LA, they went from winning convincingly by 18 points, to losing, again to GND, by 20 points. Even in chapters where they held on in 2021, their wins were not as clean as before. In Denver and North Texas, chapters they won by large margins in 2019, they only tied for first place with voters who had no alignment at all. The only chapter where they grew their share of ballots was, ironically, Triangle DSA, home chapter of one of their incumbent candidates, Kevin Richardson, who barely lost his NPC seat.  Additionally, six chapters that supported SMC in 2019 did not cast an NPC ballot in 2021: Heart of the Valley, Honolulu, Kansas City, Lansing, Tucson, and Ventura County.

Looking at this data may lead one to dwell on the factionalization of the delegates. However, if we look at bullet voting, the story is a bit different. In 2019, 45% of all ballots cast could be considered “strong” ballots for one faction or another, while in 2021 only 30% of all ballots cast were strongly aligned with one specific faction. The rate of ballots that didn’t rank all the available candidates also dropped from 87% in 2019 to 76%, meaning more delegates ranked all the candidates than in 2019. It’s difficult to say whether this is an indicator that delegates did not see one faction’s success or failure quite as instrumental as in 2019, or if this was simply a byproduct of their being less choices on offer overall, but it is worth noting. 

There is definitely more analysis we could do across this data.  We could look at vectors like votes by chapter size, chapter age (51 new chapters voted in the NPC election in 2021), voting trends for At-Large delegates, regional and state-level similarities, among others. We may pick some of this up in later posts. We haven’t even scratched the surface of what ideological tendencies this data might correspond to. We would like to let people draw some of their own conclusions from the data. 

We want to stress that this data is quantitative, and not qualitative. While numbers can explain the how of election results, they don’t paint the full picture as to the why or what it means. Lots of factors go into how well factions perform, from their political positions, to what chapters and regions they prioritize, to just plain organizing proficiency. Convention votes can give us only a snapshot of these dynamics, which will continue to play out in the months and years to come.